1977–1997...

and beyond

Nothing Stops It!

Of all the winning attributes of
the OpenVMS operating system,
perhaps its key success factor is its
evolutionary spirit. Some would say
OpenVMS was revolutionary. But I
would prefer to call it evolutionary
because its transition has been
peaceful and constructive.

Over a 20-year period, OpenVMS

has experienced evolution in five
arenas. First, it evolved from a system
running on some 20 printed circuit
boards to a single chip. Second, it
evolved from being proprietary to
open. Third, it evolved from running
on CISC-based VAX to RISC-based
Alpha systems. Fourth, VMS evolved
from being primarily a technical oper-
ating system, to a commercial operat-
ing system, to a high availability
mission-critical commercial operating
system. And fifth, VMS evolved
from time-sharing to a workstation
environment, to a client/server
computing style environment.

The hardware has experienced a

similar evolution. Just as the 16-bit
PDP systems laid the groundwork
for the VAX platform, VAX laid the
groundwork for Alpha—the industry’s
leading 64-bit systems. While
the platforms have grown and
changed, the success continues.

Today, OpenVMS is the most

flexible and adaptable operating
system on the planet. What start-
ed out as the concept of ‘Starlet’
in 1975 is moving into ‘Galaxy’
for the 21st century. And like
the universe, there is no end
in sight.

—Jesse Lipcon
Vice President of UNIX and
OpenVMS Systems Business Unit
TABLE OF CONTENTS

CHAPTER I Changing the Face of Computing 4

CHAPTER II Setting the Stage 6

CHAPTER III VAX Hardware Development 10

CHAPTER IV VMS Software Development 14

CHAPTER V Market Acceptance—Beyond Expectations 20

CHAPTER VI Moving into Commercial Markets 24

CHAPTER VII Networking 26

CHAPTER VIII The Second VAX Generation 28

CHAPTER IX Putting the VAX on a Chip 32

The Family Album

CHAPTER X Building the Bridge to Alpha 38

CHAPTER XI Alpha—The 64-bit Breakthrough 42

CHAPTER XII Inaugurating the King of Clusters 46

CHAPTER XIII OpenVMS Today 50

CHAPTER XIV Serving Customers Worldwide 52

CHAPTER XV The Affinity Program 56

CHAPTER XVI Vision of the Future 58

Major Releases of VMS and OpenVMS 60

VAX/VMS at a Glance: 20-year Timeline 62

1
In a world where computer technology becomes outdated in
three to five years, a technology that is still going strong
after 20 years deserves a round of applause. That
technology is VAX and VMS. In this, the 40th year of
DIGITAL and 20th year of VAX and VMS, DIGITAL is
reflecting on a computer platform that made history during
the last two decades of this century—and is moving forward
into the 21st century without missing a beat.

Celebrating 20 years of success, the VAX family of

computers and the OpenVMS operating system remain
the backbone of computer systems in many organizations.
OpenVMS systems have become an industry standard
in reliability, scalability, data integrity, and continuous
computing 24 hours a day, 365 days a year.

This commemorative book is intended to provide a

behind-the-scenes look at the strategy, challenges,
and people that created this globally-acknowledged
engineering marvel. It is not intended to be an
in-depth study of the technology of the platform.
Rather it is intended to celebrate a system that has
indeed become the Energizer Bunny of computers.
It just keeps on going. And going. And going.
 ACKNOWLEDGEMENTS

Editors: Content Providers: Matti Patavi

Jim Rainville Brian Allison
Karen Howard Patty Anklam Mark Plant
Paul Beck Robert Porras
Writer: Gordon Bell Jean Proulx
Kathie Peck Richard Bishop John Rando
Tom Cafarella Robert Ryan
Design: Rick Casabona Joe Scala
Kathy Nardini, Doherty Bruce Claflin Terry Shannon
Communications Wally Cole Maurice Steinman
Harry Copperman Bob Stewart
Research: Chris Christiansen Mark Stiles
Jim Rainville Marion Dancy Bill Strecker
Cynthia Carlson Jan Darden Bob Supnick
Ed Yee Stu Davidson Wendy Vogel
Ann Hewitt Scott Davis Bob Willard
Bill Demmer Ed Yee
Review Team: Sas Dervasula Steve Zalewski
Paul Bergeron Richard Doucette
Peter Bernard Elliot Drayton
Jeff Borkowski Jim Evans
Mary Ellen Fortier Dave Fenwick
David Foulcher William Fischer
Andy Goldstein Mary Ellen Fortier
Mark Gorham Sue Gault
Jane Heaney Andy Goldstein
Jackie Jones Roger Gourd
Karen Leonard Clair Grant
Steve Stebulis Heather Kane
Ed Yee Forrest Kenney
Veli Kozkko
Photography: Jim Krycka
Nancy Strader, Kim Leavitt
Corporate Photo Library Jud Leonard
Dick Willett, Forefront Rich Lewis
Magazine Steve Lionel
Jesse Lipcon
Web Production: Daryl Long
Jim Keenan Ernie Lyford
Rich Marcello
Dan Marshall
Wes Melling
Kathy Morse
Larry Narhi
Ken Olsen
CHAPTER I Changing the Face of Computing

After more than 300 man-years of intensive development, DIGITAL

announced its first 32-bit computer system—the VAX-11/780 and
its companion operating system, VMS—at the Annual Meeting of
Shareholders on October 25, 1977. Because of its 32-bit technology,
the VAX system represented a new milestone for DIGITAL and was
heralded as a major breakthrough in the computer industry.

The VAX platform and VMS operating system were unveiled by the
President and founder of DIGITAL, Ken Olsen. The new product was
showcased with a clear plastic front so that the audience could see the
CPU, cache, translation buffer, and other integral parts of the machine.
The VAX system was demonstrated running a Scrabble program—
astonishing spectators by winning the match against a human.
The winning play was the word “sensibly,” taking the 50-point,
seven-letter bonus and scoring a total of 127 points.

Over the next decade, VAX and VMS products were destined
to change the way people used computers—and catapulted
DIGITAL into a position as one of the world’s top computer
manufacturers.

Setting the sights high

The VAX system was designed to meet several key objectives.
First, it was based on a revolutionary 32-bit architecture. Second,
it solved many of the problems associated with earlier computer
technologies. Third, it was designed to be useful to the largest
possible number of users in diverse markets, and to offer
DIGITAL customers a seamless transition from earlier
product architectures. And finally—in direct opposition to
the questionable strategy of planned obsolescence—the
VAX system was designed to last between 15 and 20 years.
An impressive list of requirements, to be sure. But VAX
and VMS met them all.

“The best of what we’ve learned about interac-

tive computers in our first 20 years has gone
into this machine. We have spent more than
300 man-years of intensive engineering effort
in its development, and during that time
I have sensed more excitement and
enthusiasm among the developers of
VAX than I remember seeing at any other
time in the short history of DIGITAL.”
—Ken Olsen
Founder, Digital Equipment Corporation
October 25, 1977

“Jesse Lipcon continually reiterated, ‘Our top three goals are time to
market, time to market, time to market.’ At one point I said, ‘Wait a
minute, Jesse, what about quality?’ Without missing a beat, Jesse replied,
‘Quality isn’t a goal, it’s a given.’”
—Jay Nichols
Computer Special Systems, Manager of Engineering

4
 Moving from 16-bit to 32-bit computing.

A total systems focus

From its inception, the VAX development program had a total system focus,
encompassing groups from hardware and software engineering, support,
product management, documentation, and manufacturing. More than 1,000
people in the corporation worked on the first VAX and VMS system, in some
capacity, on an extremely aggressive schedule. Without question, the work
produced within the limited time frame exceeded all expectations.

Hardware meets software at the drawing table

VAX and VMS made engineering history by being the first interactive
computer architecture in which the hardware system and software system
were designed together from the ground up. This was a novel approach to “In our spare time, Stan Rabinowitz
designing a computer architecture, where hardware and software teams and I wrote a Scrabble program for
worked jointly and altered their designs in consideration of each other’s the PDP-11. As soon as the VAX was
requirements. The result of this united engineering effort was a tightly available, we ported it over to the
integrated system that provided unprecedented reliability, flexibility, VAX. Ken insisted that we demon-
scalability, and data integrity. In short, bullet-proof computing. strate the Scrabble program at the
announcement. I ran the Scrabble
The first VAX system demonstrated a major industry breakthrough by program—pitting it against a
providing the functionality, capacity, and performance of a mainframe— human being—and the VAX
coupled with the interactive capabilities, flexibility, and price/performance demolished him. Then Ken stood
of a minicomputer. up and said, ‘It’s time for the big
league games. This isn’t Tic Tac Toe,
this is Scrabble!’”
—Richie Lary
Corporate Engineering Consultant

5
CHAPTER II Setting the Stage

The pre-VAX years

The first VAX computer was introduced as DIGITAL celebrated its 20th
anniversary. The company—founded by Ken Olsen, Harlan Anderson, and
Stan Olsen in 1957 with an initial capital investment of $70,000—began as
a small module manufacturer in a corner of a sprawling mill complex in
Maynard, Massachusetts, a small town 30 miles west of Boston.

Moving from printed circuit modules to computers

While DIGITAL initially produced printed circuit logic
modules, the company’s real mission was to bring
computing to the people. In its second year, DIGITAL
made the transition to computers and in 1959 introduced its
first computer—the PDP-1. During the 1960s, the company
rolled out a family of PDP computers, each more powerful
than its predecessor. Early on, innovation and engineering
excellence were hallmarks that have characterized
DIGITAL products throughout its entire history.

Peer-to-peer networking—the birth of distributed

computing
The company’s first computers were stand-alone systems.
But in the early 1970s, DIGITAL pioneered peer-to-peer
computer networking with the introduction of its first suc-
cessful networking software product, DECnet. Networking
allowed customers to connect many minicomputers and
share a common database of information. This approach to
computing launched the concept of distributed computing.
It was a novel approach because at the time mainframes
and stand-alone minicomputers were the only computing
game in town.

Front cover of Fortune magazine, October 1986. Features Up to that point, computers did not talk to each other.
story on Ken Olsen and Digital Equipment Corporation. Moving information from system to system involved using
slow magtape and sneaker net. Distributed computing
offered the advantage of flexibility and connectivity. Now
information could be moved across computer rooms and
later across the country almost instantly—making the DIGITAL goal of
bringing computing power to the people who needed the information a
reality. Technical customers embraced the interactive, accessible nature of
these new minicomputers, and DIGITAL began to flourish with PDP
systems selling in the tens of thousands.

“Gordon Bell’s vision was the primary driver behind the entire VAX family.
And I think its success was due to his vision.”
—Bill Demmer
Former VP, Computer Systems Group

6
 Digital Equipment Corporation headquarters
in Maynard, Massachusetts, with the famous
clocktower.

The beloved Mill

The physical environment in which VAX and VMS was cre-
ated was very much in keeping with its New England her-
itage. “The Mill”—a set of brick buildings in the center of
Maynard, Massachusetts—exemplifies two old New
England traditions. One is the classic mill town pattern
with the development of an industry and the growth of a
community around it. The other is the thrifty Yankee
“make do” principle—it’s better to “make do” with what
you have if it’s still useful, rather than abandon it and buy From textiles to computers
something new and expensive. In 1953, ten businessmen from nearby Worcester bought
the mill and leased space to tenants. One
The original mill site on the Assabet River was once part of of the companies attracted by the affordable space was
the town of Sudbury; the opposite bank belonged to the Digital Equipment Corporation, which started
town of Stow. The present town, incorporated in 1871, was operations in 8,680 square feet in the mill in 1957.
named for the man most responsible for its development,
Amory Maynard. At the age of 16, Maynard ran his own DIGITAL grew so fast that within 17 years, it bought and
sawmill business and later went into partnership with a expanded into the whole mill complex. Inside,
carpet manufacturer. They dammed the river to form a old paint was removed from the walls, exposing large
millpond to provide power for a new mill, which opened in areas of the original brickwork. Pipes were painted
1847. in bright colors, in contrast to the massive beams
The clock tower and columns. The large interior once used for textile
After Amory Maynard died in 1890, his son Lorenzo built machinery became filled with modular office cubicles—
the Mill’s famous clock tower in memory of his father. The offering flexibility to meet the company’s changing
clock’s four faces, each nine feet in diameter, are requirements. DIGITAL left the exterior of the buildings
mechanically controlled by a small timer inside the tower. largely unaltered, but cleaned up the Assabet River, which
DIGITAL never electrified the timer nor the bell mecha- once was colored with the residue from the
nism. To this day, someone has to climb the 120 steps mill’s dyeing plant.
once a week to wind the clock: 90 turns for the timer and
330 turns for the striker. When the employees first moved in, the floors were wavy
and soaked with lanolin from the wool processing days—
In 1899, the American Woolen Company, an industrial which would eat right through crepe-soled shoes.
giant, bought the Assabet Mills and added most of the
existing structures. The biggest section was Building 5, While the engineers started designing the VAX and VMS
which was 610 feet long and contained more looms than architecture, the floors were being refinished, so they had
any other woolen mill in the world. to live with the ambient music of hammering on the floors
above. Then the floors were sanded and polyurethaned.
Over the next 50 years, the Assabet Mills survived two During the sanding, the engineers were given plastic
world wars and the Depression. But when peace returned, sheets to cover their desks and equipment each night.
the Assabet Mills were shut down entirely in 1950. Like Before the roof was replaced, rivers of water flowed
many New England mills, it succumbed to a combination through parts of the mill when it rained.
of Southern and foreign competition and the growing use
of synthetic fibers.

7
Industry trends of the time
During the mid 1970s, industry trends included interac-
tive computing and networking. Corporations were
discovering that distributed computing was a viable
alternative to the mainframe batch environment.
Distributed computing allowed a company to
decentralize information and put it in the hands of
decision makers—an idea which dovetailed perfectly
with Ken Olsen’s original goal of giving computer
power to the people. PDP-11/70—moving from 16-bit to 32-bit computing.

Due to industry-wide technological advances, more computing power

could be packaged in every square centimeter of space for less money.
As a result, computer systems could be made smaller and more affordable,
and provide functionality previously found only in large mainframes.

DIGITAL drove the industry trend of building increasingly more

powerful—but less expensive and physically smaller—computer systems.

DIGITAL had developed tools to meet the opportunities created by these

industry trends: the interactive minicomputer, DECnet, more powerful
and easier-to-use software, a volume manufacturing capability, and
financial applications. However, corporations were being constrained by
the limitation in the addressing range of 16-bit computer architectures—
a bottleneck the industry needed to address.
“I would say probably the most
significant thing that DIGITAL has
Outgrowing 16-bits
done is make computing available
As early as 1974, DIGITAL recognized the limitations of its 16-bit PDP
to the masses. Instead of the high
architecture—especially for such tasks as writing large programs and
priests in the white robes behind the
manipulating large amounts of data often needed in scientific, engineering,
glass walls, DIGITAL brought com-
and business data processing applications. The company realized there
puting out of the glass house and
was a need for a new architecture that would be compatible with PDP
made it affordable and acceptable
systems, but would have larger addressing capabilities and enough power
to the mainstream.”
to meet the computing needs of the future.
—Terry Shannon
Publisher, Shannon Knows DEC

8
32-bit computing: the next logical step
Some users were also beginning to feel hampered by the limitations of
16-bit computing. They were becoming frustrated that large programs had
to be broken up into smaller pieces in order to run on their computers.
The 16-bit addressing was hindering progress, and finding a solution
became eminent. Extending the addressing architecture to 32 bits could
supply the power needed to solve this problem. Other computer companies
also recognized the shortcomings of 16-bit addressing and had begun
working on 32-bit systems, which seemed to be the next logical step in
computer development.

Considering the alternatives

DIGITAL was intent on maintaining its lead in the minicomputer industry
and knew that extending the addressing architecture was critical. The
company considered many approaches to extending the addressing that
built on its current line of products, the PDP family and the DECsystem-10.
The company also knew that customers not only required more
system power and memory, but wanted more economical systems that
would be compatible with their PDP-11
family of processors, peripherals, and software.

In late 1973, DIGITAL began development of the

PDP-11/70—an extension of the basic architecture of
the PDP-11 family—in an attempt to solve the memory-
addressing problem. The PDP-11/70 had a larger memory
capacity (up to 4MB), but using it with the 16-bit PDP-11
software architecture was cumbersome. The development
team had the choice of finding a way to continue to “brute
force extend” the PDP-11 family architecture, or to create
something new.

VAX-11/750

9
CHAPTER III VAX Hardware Development

Creating a whole new architecture

In March of 1975, a small aggressive development task force was formed
to propose a 32-bit PDP-11 architecture. The team included representation
from marketing, systems architecture, software, and hardware.

“It’s the addressing capability that The project included three phases. During Phase I, the team produced a
has probably been, over history, document that encompassed a business plan, system structure, build plan,
the major driving force behind new project evaluation criteria, the relation to longer term product development,
computer architectures. It’s what software, and the alternatives considered. In Phase II, they produced the
initially brought about the whole project schedule. Phase III was the implementation of the program.
notion of the VAX computer.”
The planned ship date for the new hardware system was for 18 to
—Bill Demmer
20 months from the start date. The overriding concern was getting to
Former VP, Computer Systems Group
market quickly with a 32-bit system to satisfy customers’ need for more
computing power.

Setting celestial sights

Initially, the VAX and VMS development team used the code name “Star”
for the hardware and “Starlet” for the operating system. Thus began the
celestial code naming of hardware and software. Plans for a new family
of 32-bit systems had already been drawn. Over time, about 40 engineers
worked on the hardware development team. Everyone involved expected
this new class of computers to
propel DIGITAL to the forefront
of technology—and the air was
charged with excitement and
enthusiasm.

Planning the project

Gordon Bell, VP of Engineering, was
the primary driver behind the new
direction for DIGITAL. Bell drew up
plans for a new system that would
extend the addressing scheme used
on the PDP-11.

While the initial code name for

the new system was Star, it soon
became known internally as VAX,
an acronym for Virtual Address
eXtension. When the product was
announced, the company added the
number 11 to the name VAX to show
customers that the new system was
compatible with the PDP-11.

Above, VAX-11/780 Announcement, from left to right: Gordon Bell, Richie Lary, Steve Rothman,
Bill Strecker, Dave Rogers, Dave Cutler, and Bill Demmer.

10
 Ken Olsen powers up first VAX. The first time
the original VAX breadboard (prototype)
was powered up, Ken Olsen almost burned
Bill Demmer was the project manager for the VAX project and assembled his hand on the power supply.
three design teams to work on the new architecture. VAX A Team developed
the concept design plan; VAX B Team initiated some of the architectural
extensions and provided ongoing review of the design specification and
project plan; VAX C Team reviewed and approved the final project plan and
design specification.

Working the plan

The VAX and VMS kick-off meeting took place in April 1975. The task
force—addressing such items as instruction set extensions, multiprocessing,
and process structures—closeted themselves away to discuss their options.
Their goal was to make the least possible changes in the PDP-11, and still
extend it to have a larger virtual address base.

The fundamental questions were, “Can the PDP-11 architecture have its
addressing structure expanded to achieve the goals of transparency to
the user? Can this expanded architecture provide a long-term competitive
cost/performance implementation similar in style and structure to the base
architecture?” It was not long before the team realized that both these goals
could not be met completely, and so they made the decision to develop a
completely new architecture.

All team members agreed that the new system would have to be culturally
compatible with the PDP-11 and it would have to maintain the same look
and feel as its successful predecessor.

An architecture evolved that cleanly solved the fundamental limitations

of the PDP-11. The team developed an implementation plan to overlay
both the extended architecture and the basic PDP-11 architecture, thus
permitting the new system to appear to be an extended model of the PDP-11
family. This would help to achieve one of the major goals—allowing cus-
tomers to capitalize on their investment in PDP-11 and grow their systems.

In addition to expanding the address space and ensuring PDP-11 compati-

bility, another goal was to create an architecture that would support user
requirements for 15 to 20 years.

11
 Evolution of the architecture
For almost a year, the VAX devel- Machines would range from desktop
opment and review teams worked to enterprise-wide systems. The
back and forth on the VAX archi- goal was to establish a single VAX
tecture. After four versions, the distributed computing architecture
proposed architecture was found that would run the same operating
to be too complex, too expensive, system. A related goal was that VAX
and too complicated to execute. products would one day provide a
price range span of 1000:1.
The company formed a group that
became known as “The Blue Ribbon The competition offered larger
Committee” that included three machines for upward migration.
hardware engineers: Bill Strecker, Since these machines did not run
Richie Lary, and Steve Rothman, the same code, the code had to
and three software engineers: Dave be recompiled from the smaller
“VAX was the project name—Virtual Cutler, Dick Hustvedt, and Peter machine in order to run. The
Address eXtension—but it was never Lipman. They simplified the earlier DIGITAL single-architecture
meant to be the product name. design and created a plan that strategy equated to cost savings
When it came time to choose a would be possible to execute. Key for customers and simplified their
name, we thought PDP-what? Then modifications included drastic sim- computing environments.
some marketing specialist said there plifications to the highly complex
are two attributes that are really memory management design and In addition, a single architecture
important in a name, if you want it to the process scheduling of the pro- enabled the building of network and
be memorable. One is that it be posed system. The simplified archi- distributed processing structures.
short and pronounceable and that it tecture, the fifth design evolution of
have an X in it, because Xs are rare the VAX system, was perfected and Implementing the VAX hardware
letters, so they catch your eye. accepted in April of 1976—exactly Once the plan was accepted, the
According to that theory, we had the a year after the design work began. VAX and VMS project entailed
best name sitting right in front of us: several months of laying out the
The VAX strategy basic design for the architecture,
VAX.”
Simplicity was the essence of the followed by nine months of filling
—Peter Conklin, VAX strategy. The VAX strategy pro- it in. The plan was executed by
VMS Engineering Manager vided for a set of homogeneous, dis- two separate hardware engineering
tributed computing system products teams. One used existing technology
that would allow users to interface, to design what eventually became
store information, and compute on the VAX-11/780; the other team
any of the products—without having developed a new VAX chip tech-
to reprogram their applications. nology through the DIGITAL
fledgling semiconductor group

“I don’t think many people ever get the kind of opportunity we did. We had
good people, and we grew into a great team. We had lots of differences, but
we sorted them out and built what was expected.”
—Roger Gourd
Software Engineering Manager

12
which became the VAX-11/750.
Memory and CPU design
In planning the memory design,
there was a question of what size
memory and how many bits were
needed. Trade-off decisions were
made between achieving the best
performance and optimizing the
number of bits used from a cost per-
spective. The VAX-11/780 memory
design was the first in which error-
correction and detection code (ECC)
was designed into the system. The
semiconductor DRAM (Dynamic VAX-11/780 – moving from 16-bit to
Random Access Memory) was 32-bit computing.
susceptible to soft errors. In order
to protect the system from memory
loss or changing information, it was
necessary to store the information in
the memory using some code with
additional bits of memory. In the
unlikely event that one of the bits
changed, the memory system could
reconstruct the code, know which
VAX...when you care enough to
one changed, and correct the
steal the very best
problem.
During the cold war, VAX sys-
tems could not be sold behind Actual Russian words translated: VAX... when
With virtual memory, memory you care enough to steal the very best.
the Iron Curtain. Recognizing
space no longer had to be in the
superior technology, technical
system’s internal memory all at
people cloned VAX systems in
once. Instead, the whole program
Russia, Hungary, and China. After
sat on a disk, while the operating
learning that VAX systems were being
system moved pieces in and out of
cloned, DIGITAL had the following
internal memory as needed.
words etched on the CVAX chip,
Because the first VAX systems had
“VAX...when you care enough
very little internal memory, this was
to steal the very best.”
important. Relative to internal mem-
ory, disk memory was economical.
Virtual memory also allowed pro-
grams that were too large to fit com-
pletely into memory to run parts at
a time—which was not possible on
systems that did not do virtual
addressing.
“In the early 1980s, we were designing computers so complex, our
engineering processes couldn’t keep up with them. We discovered we
had to use the latest VAX to simulate the new one we were building.
Building VAXes on VAXes—our first computers became tools for building
the next generation of VAXes.”
—Bill Strecker
Chief Technical Officer, VP, CST

13
CHAPTER IV VMS Software Development

With the VAX hardware development underway, the software

development—code named Starlet—began a few months later in June of
1975. Roger Gourd led the project and software engineers Dave Cutler, Dick
Hustvedt, and Peter Lipman were technical project leaders, each responsi-
ble for a different part of the operating system.

VMS project plan

The Starlet project plan was to create a totally new operating system for
the Star family of processors. The plans called for a high-performance
multiprocessing system that could be extended to support many different
environments. Just as the hardware was designed to be culturally
compatible with the PDP-11, Starlet was designed to augment the hardware
compatibility by providing compatibility with the existing operating system,
RSX-11M.

The short-term goal was to build an operating system nucleus for the first
customer shipment of VAX systems. It would have sufficient functionality
to be competitive, but would also provide a base that could be extended
and subsetted over time for a variety of DIGITAL markets. Long-term
goals for the project included quality, performance, reliability, availability,
serviceability, reduced support costs, and lower development and mainte-
nance costs. The main focus was to support high-performance applications,
such as real-time and transaction processing.

VMS V1 software development team.

14
Putting it in writing Close cooperation between the hard-
From the beginning, the software ware and software engineers also
team considered documentation to helped work out potential software
be a significant part of the project. problems, and ultimately created
The first technical writer, Sue Gault, a hardware and software system
attended design meetings with the that was tightly integrated. It meant
software development team and the difference between being
helped them write the Starlet designed-in rather than added-on
Working Design Paper. This docu- later. Many hardware factors
ment contained an in-depth techni- changed as a result of the software
cal description of the operating sys- development work.
tem. Since this project was defined
as building a system of hardware The VAX and VMS development
and software together, it was more team recognized that an early
complex in scope than any DIGITAL hardware implementation was
Creativity with discipline
project to date. critical to developing the software
From its inception, DIGITAL realized
concurrently with the real
the importance of creativity, and
Through the exercise of writing, the hardware.
sought to create an environment
engineers received input from the
in which individual creativity would
documentation writers and were Accordingly, a team of hardware
thrive within a disciplined environ-
able to troubleshoot potential prob- engineers built a system called
ment. The strategy was to hire
lems. Ideas had to be expressed the “hardware simulator.”
talented people and empower them
clearly enough to be written in the Constructed of PDP-11/70
to develop plans, have those plans
specs. This method helped to re- components, some custom logic
reviewed and approved, and to
solve assumptions and potential dif- boards, and a lot of firmware,
accept ownership of the project. The
ferences of opinion. The design doc- it provided a quick first imple-
company believed that people would
ument also served to keep the rest of mentation for the VAX platform.
be most productive when they had to
the company informed about the The VMS team designed and
meet milestones and stay within a
VMS project as it was made public— tested all the operating system
budget—this is where the discipline
contributing to the overwhelming software on that simulator. It ran
factor came
support and enthusiasm throughout 10 to 20 times slower than the
in. Each product line group was
the company for the new project. actual system was to run, but it
responsible for meeting its plan with-
enabled the development team
in time and budget constraints.
Working in tandem to work and develop the software
In order to ensure tight integration on the system as it was being
between the software and hardware, designed.
several software programmers at-
tended the VAX design committee Breadboard on wheels
meetings and contributed to the The VAX-11/780 engineering team
hardware design from a software first built a machine called the
perspective. breadboard and placed it on a large
metal cart. All the circuit boards
were created with wire wrap; while

“Software development is very creative, very individual. We want to give the

engineers the freedom to work independently, to work together, and to do the
things they want to do.”
—Bill Heffner
VP of Software Engineering

15
 the power supplies sat loose on the lower shelf of the cart. The software
development team ran a time-shared VMS system on the breadboard for a
short time, because it became available around the time VMS had evolved
sufficiently to support multiple users. However, the breadboard was not
All work and no play? entirely reliable because the operating speed was pushing the
Not at DIGITAL! limits of what could be handled with wire-wrap construction.
Over the years, the VMS engineers
laughed together as well as worked The breadboard was replaced by the first VAX-11/780 etched prototype.
together. And so there were a whole The prototype was the first machine built with “real” parts—the real frame,
series of practical jokes that were power supplies, etch circuit boards, etc. The only thing lacking was the
played. There were some guidelines: external cabinet. This prototype was not replaced with a production
You couldn’t prevent people from get- machine until well after VMS and the VAX-11/780 shipped. The prototype
ting work done. You weren’t allowed continued to be used for years for stand-alone testing. Systems developed
to do anything that would harm the after the VAX-11/780 never went through the breadboard stage, but rather
system or lose a day’s work. But any- went directly to real etched circuit boards, after extensive simulation.
thing else was fair game.
The software developers used the prototype to do their work—which provided
Dave Cutler started the first VMS a closed loop of using what they were building. This strategy of using the
April Fool’s jokes. One year, Andy software system to do the design work helped them to pinpoint potential
Goldstein replaced the line printer dri- problems as they progressed.
ver so that everything printed
out backwards. On another April 1,
the entire system message file
was replaced with joke messages—
including ones like “File not found.
Where did you leave it?”

Once VMS engineer Trevor Porter

went back home to Australia on
vacation. When he returned, fellow
engineer Andy Goldstein had bolted
a panel in place where the cube
“door” was. Trevor walked to his
office, observed the situation, turned
to Andy and said, “All right, where’s
the spanner?”

VMS engineers Dick Hustvedt and Ben Schreiber.

“One of the VMS group’s philosophies was that we lived

on the software that we were writing. Because if it wasn’t
good enough for us, then it wasn’t good enough yet for
our customers.”
—Kathy Morse
VMS Engineer

16
 Who’s got the red flag?
The software build environment process, which is what transforms the
software source code into a runnable system, allowed only one person
at a time to do a build. If two people tried to do a build at the same time,
they would overwrite each other and produce nothing useable. The
engineers—who often worked in an intense, heads-down mode—had
“Roger Gourd passed around the no way of knowing if another engineer was working on a build. It was
book The Mythical Man-Month by inevitable that early on, two engineers in adjacent offices would try to
Fred Brooks and almost all the team do builds concurrently, thus destroying each other’s work.
members read it. Most of us already
had one operating system under our Being a creative team, they came up with a creative solution. As a
belt, so Brooks’ discussion of the mechanism for determining who was working on builds, a red flag with
‘second system effect’ struck home. a magnetic holder was put up in the cubicle of the person using the
The ‘second system effect’ results simulator.
from each engineer wanting to fix all
the mistakes and shortcomings of It was usually referred to as “the mutex,” in reference to a commonly
their first system. Left unchecked, used software synchronization mechanism. If an engineer wanted to do
the second system effect can cause a build, he or she found the flag and asked its current owner “Can I have
runaway complexity that can be dis- the mutex?” and it would be theirs as long as the flag holder wasn’t in the
astrous for middle of a build.
software quality and schedule.
A new term entered the program- Developing tools on the fly
mers’ lexicon—‘Creeping elegance’—a Out of necessity, the team developed many of their own tools as the
process in which a design is succes- project progressed. For performance evaluation, the VMS engineers built
sively refined to be increasingly com- the performance monitor tool and then used the tool to measure system
plete, eventually yielding a result that performance. One part of the monitor was a separate computer system
collapses because of its size and running on a PDP-11 that could act as a time-sharing workload. Using that,
complexity. The entire software team the engineers measured VMS on a number of different multi-user work-
was very conscious of maintaining loads to see how it performed for time-sharing.
the balance between producing a
functional, high quality product and Using what you’re building
staying on schedule.” There was a lot of back and forth communication between the hardware
—Andy Goldstein and the software engineers. The writers were also using the software,
VMS Engineer on original which provided a good closed loop process. And that was the philosophy
development team behind it—to use it and debug it as the project moved forward.

Ensuring compatibility
The development systems for VMS were housed in one large computer
room; most of it was taken up by a huge dual-processor DECsystem-10
and a PDP-11/70 was shoehorned into one side of the room. Much of the
first version of VMS was written in Macro and the rest in Bliss. Macro
development was done strictly on the PDP-11, using a cross-assembler.
The assembler object modules were then linked into executables on the
PDP-11 and written to a disk which the VMS engineers would then carry
over to the VAX system in the next room.

17
VMS documentation set.
The DEC-10 was used to compile the VMS modules written in Bliss
because at the time the Bliss compiler only ran on a DEC-10. The Bliss
code had to be transported by tape to the PDP-11 to be linked.

This process of writing programs initially required a great deal of time and
effort. However, the new virtual memory operating system was built in a
That’s not an relatively short time by any current standards.
abandoned car—
it’s a VMS engineer’s car The VMS system kernel and related critical function were written in “native
“People worked a lot of overtime dur- mode” using the new VAX instruction set. However, many utility functions
ing the creation of VMS. At one point, were simply ported from the RSX-11 operating system and so ran in
we hired an engineer from California, “compatibility mode”—the PDP-11 emulation mode. Besides speeding up
Ralph Weber. For the first week he the implementation of these functions on VMS, this approach provided an
had a rental car and was living in a effective live test of the VAX platform and VMS operating system
hotel. He got there so early that he compatibility features.
parked in exactly the same spot
every morning, and he stayed late. The virtual memory system software provided greater functionality than
After a week, a security guard had ever been seen before in a minicomputer. VAX and VMS also supported
thought the car had been aban- networking capabilities as well as compatibility with PDP-11 thus enabling
doned and called the car rental customers running PDP-11 programs to migrate their applications to the
place to come and collect it. new VAX and VMS systems quickly and easily.
That night Ralph went to leave, and
his car was gone. So he ran into the The VMS strategy
security room shouting, ‘My rental The VMS software strategy was based on developing a single VMS
car’s been stolen!’ They started to operating system that would span the product range from low-end to
call the police and then, luckily, high-end. VMS would offer full mainframe capabilities allowing concurrent
another security guard came in and batch processing, transaction processing, time-sharing, and limited
said, ‘No, no, we had that one towed real-time processing.
today because it’s been there a week
and we thought it had been aban- This single operating system strategy behind VMS was a reaction to the
doned.’” multiple operating systems of the PDP-11:
—Kathy Morse • RT-11 for real-time and laboratory work
VMS Engineer • RSTS-11 for educational and small commercial time-sharing
• RSX-11 for industrial and manufacturing control
• MUMPS-11 for the medical systems market
• DOS-11, the original PDP-11 operating system, largely superseded
by the above.

18
VMS Version 3 release party
on Cape Cod.

Debugging in the Blizzard of ‘78

On the first evening of the blizzard,
Andy Goldstein was working late on
the new VMS file structure. If he
couldn’t make it home, he wasn’t
worried. Hank Levy lived across the
While each of the PDP-11 operating systems was targeted to a particular road from the Mill, so anybody from
market segment, there were a lot of cross-over sales. At the same time, the VMS who was really stuck would just
multiple operating systems with incompatible interfaces diluted the system pound on his door and sleep on his
base for applications. Any application might have to be implemented in couch.
multiple versions to run on a large number of systems.
“I hit a bad directory error and said,
Oh my God, I’ve got a bug
Therefore, the strategy with VMS was to have a single operating system
in the file system. I was trying to col-
that would be sufficiently flexible, powerful, and efficient to address most
lect data on this, but the snow was
of the PDP-11 target markets.
getting deeper and we lost power.
The whole state was closed for the
Betting the business on VAX and VMS
next week, but I drove to the Mill and
Prior to developing the VAX system and VMS operating system, DIGITAL
talked my way inside.
operated according to a multi-product line environment. However, in 1978,
I powered up the machine, got
DIGITAL adopted a vision called The VAX Strategy which would guide the
dumps of the failed directory,
company through the next decade. Although DIGITAL would continue
and took them home with me.
development on the PDP-11 and DECsystem-10, the company’s main
direction would be on VAX development. “I called Richie Lary—who lived
across town from me—and said,
The VAX and VMS strategy led to a consistent message from DIGITAL: ‘Richie, I think there’s a bug in the
“One platform, one operating system, one network.” Simply put, DIGITAL microcode.’ And he said, ‘Why don’t
decided to bet the business on VAX and VMS—and VAX and VMS business you come over. I’ve got the microc-
began to skyrocket. ode listings here.’I walked through
the snow over to his house. Richie
fished a six-inch binder out from
under his bed and we went through
it, and sure enough, we found the
“As the technical writer, my belief was that the technical writer is the bug and
advocate for the customer. So I always put myself in the shoes of someone fixed it.”
who is trying to learn how to use the system, and wrote the documentation
accordingly. —Andy Goldstein
VMS Engineer
“The VMS Documentation Group grew from five people in 1977 to 45 in
1987, and the documentation set grew from 9,000 to 20,000 pages. It was
a massive effort.”
—Patti Anklam
Technical Documentation Writer

19
CHAPTER V Market Acceptance—Beyond Expectations

Rolling out the first VAX and VMS systems

Roughly 18 months after the design team sat down to execute their plan
for the new interactive architecture, the first machine rolled off the
manufacturing floor and into a customer site.

The first VAX-11/780 was installed at Carnegie Mellon University and was
released to more than 50 customers. In 1978, the VAX-11/780 became
accepted internationally with installations at CERN in Switzerland and the
Max Planck Institute in Germany.

A major industry contribution

In October of 1977, DIGITAL made a significant contribution to the industry
by announcing both a new architecture hardware product and a new archi-
tecture-based operating system. One
of the primary advances that the
VAX architecture brought to com-
puting was that it had a plan for the
intercommunication of computers at
the architectural level. DIGITAL had
not only engineered the capability
for computers to talk to computers
with homogeneous existing archi-
tectures, but had planned for a com-
plete range of computer systems—
from the personal workstation
level up to the high-performance
systems—all having a homogeneous
architecture.

President Ronald Reagan visits a DIGITAL

VAX manufacturing facility with DIGITAL
President/CEO, Ken Olsen.

20
 Overcoming resistance to change
When VAX and VMS systems became available in 1978, customers were just
beginning to understand the need for a 32-bit architecture. Analyst reports
published after the introduction of VAX and VMS system discussed the sig-
nificance of the 32-bit architecture.

While some forward-thinking customers embraced the advantages of the

Woods meetings 32-bit architecture—especially in specialized scientific applications—many
In 1983, DIGITAL began to hold were still satisfied with their current 16-bit architectures and didn’t think
day-long, off-site meetings. Initially the larger addressing space was necessary. Resistance to change
these meetings were held at Ken is always an obstacle in introducing new ideas, and the VAX platform
Olsen’s cottage deep in the woods certainly represented a change for customers.
of Maine. Soon, these off-site strat-
egy meetings became known Migrating from PDP to VAX
throughout the company as woods As customers saw how efficiently VAX and VMS worked in their environ-
meetings—regardless of where they ments, acceptance for the new system grew overwhelmingly positive.
were held. Organizations suddenly proclaimed, “We are a VAX and VMS company”
and focused all their efforts in that direction.

The VAX system drew on years of DIGITAL engineering experience in

developing the PDP family of computers. Wherever possible, the VAX
architecture took advantage of existing PDP-11 technology such as the
UNIBUS—thus allowing existing PDP-11 I/O technology and products
to be used on the VAX systems.

Thus, the new VAX system appealed to the installed base of PDP customers
because of the built-in compatibility mode—which provided an easy
migration path for moving up to the new 32-bit architecture, while still
protecting their existing PDP investment.

21
 Key success factors
32-bits at an affordable price
Although the VAX-11/780 was not the first 32-bit system on the market,
it was the earliest computer that was capable of taking on large-scale
problems at a reasonable price.

FORTRAN for the scientific world

The DIGITAL investment in VAX FORTRAN is credited with some of the
VAX and VMS architecture’s early success in the marketplace by gaining a
leadership role in the world of scientific and technical computing.

Two aspects of VAX FORTRAN contributed to the early success of the VAX.
First, the compiler produced excellent quality, fast performing code. Since it
was a very complete implementation of FORTRAN, a FORTRAN program
written for a competitor’s machine could easily be brought over and run on
a VAX system. Second, the interactive, source-level debugging allowed the
programmer to interact with the program in FORTRAN, rather than
machine language.

VAX systems became the first workhorse for numerical and scientific com-
puting, supporting such power-hungry applications as computer-aided
design, flight, operator training for nuclear and conventional fuel power
plants, power monitoring and control systems for electric utilities, and seis-
mic data reduction.

Scalability
By design, the VAX architecture was scalable, meaning that code written
on small machines would run unchanged on larger machines. This made
software development affordable, because the concept could be tested on a
small machine before making a major hardware investment. Applications
were not limited to a particular machine. Once an application was written,
it could run on any size VAX system without changes. Scalability allowed
customers to grow their VAX systems as they needed—without worrying
about their software investments. It also minimized maintenance and
support costs for software.

Connectivity
Another significant success factor was the connectivity strategy—intercon-
necting computers via networks. DIGITAL had developed DECnet in 1973,
and support for this networking software was an integral part of the VAX
strategy. The ability to connect computers gave minicomputers the power of
mainframes. Distributed computing was an emerging concept and DIGITAL
was in the leadership position.

The VAX and VMS architecture allowed for networking that was more
VAX and VMS play a part in the efficient than in other systems at the time. Networking allowed DIGITAL
Space Shuttle development. to expand the application base for VAX platform and broaden its market
base beyond the scientific and into the commercial world.

22
Software capabilities
With the increased breadth of software offerings of VMS V2.0,
DIGITAL moved into the business marketplace.

These factors—along with the extensive software library and other interac-
tive features of VMS—made it the best software development environment
in the industry. In addition, the robustness and reliability of even the early
VMS versions ensured that a customer’s programming staff spent their time
working on their programs rather than figuring out what had gone wrong
with the operating system.

In short, the expanded address space, sophistication of the operating

system, inherent networking capabilities, and affordable price were
the integral factors in the success of this new technology.

23
CHAPTER VI Moving into Commercial Markets

By 1979, the company’s sales rev- With this announcement, DIGITAL

enues topped the $2 billion mark made a commitment to the commer-
for the first time. DIGITAL was a cial marketplace and positioned the
major player in the worldwide VAX system as the flagship product
minicomputer market, and was for new commercial applications.
marketing its systems, peripherals, The company also emphasized
and software in 35 countries networking and distributed data
around the world. processing concepts as ongoing
efforts.
This success was attributed to Ken
Olsen’s original strategy of selling Expanding the family
small, easy-to-use computers to a The initial evolution of the VAX
wide range of customers. Research family was downwards in size and
scientists, accountants, banks, price. This was a very deliberate
and manufacturers alike could use strategy that was established when
these systems. the VAX architecture was first con-
ceived; the slogan was “$250K and
By the 1980s, DIGITAL had DOWN.” Even though the DEC-10
established itself as the number-two and DEC-20 systems were still going
computer company behind IBM. strong , the intent was that the VAX
At this time, VMS Version 2.0 system would provide a replacement
commercial software was intro- for high-end PDP-11 systems—with-
duced. This second generation of out encroaching on the DEC-10/20
VMS provided to the commercial business.
DIGITAL President/CEO, Ken Olsen addresses marketplace the same leadership
customers to deliver new product announce-
ment. that FORTRAN did for the scientific
world. By the spring release in 1980,
Version 2 had users at 1,400 sites.

DIGITAL spreads the word on the ease of scalability with VAX and VAXcluster systems.

24
 Increased software capabilities
The backbone of the new VMS
commercial software capabilities
was represented by six new
products:
• COBOL was the flagship product
• BASIC was interactive
and fast
Thus the first two successors to the VAX-11/780 were smaller, less expensive
• Multikey ISAM provided effective
machines. The first successor to the 780 was the VAX-11/750. The VAX-11/750
data management and
was built of semi-custom LSI logic known as gate arrays. Each gate array
was usable from all languages
chip consisted of about 400 standard logic functions. By interconnecting the
basic functions, each chip was specialized to provide the needed functions • Integrated DECnet enabled
of the 750 CPU. multi-system communication
• DATATRIEVE V2 provided
The VAX-11/730 was the third member of the VAX family, introduced in
online inquiry and retrieval
1982. The VAX-11/730 was built from off-the-shelf bit slice microprocessor
and programmed array technology. • Form Management System (FMS)
allowed for data entry and trans-
More power, please! action-oriented
Meanwhile, some customers were beginning to clamor for more powerful applications.
VAX systems. In an effort to meet this demand, DIGITAL produced the
This extended the range of
VAX-11/782. This system was built with two standard VAX-11/780 proces-
capability and power—enabling
sors using a shared memory. By supporting the VAX-11/782, VMS took its
commercial customers to distribute
first step into multiprocessing—foreshadowing the symmetric multiprocess-
their data processing more easily
ing capabilities of the VAX 6000 series years later.
and efficiently.
This system was followed by the VAX-11/785—a re-engineered VAX-11/780
that used the same design with upgraded components—which allowed the
CPU to be run at a 50% faster clock rate. Both the VAX-11/782 and
VAX-11/785 were designed to bridge the long gap between the 780
and the 8600.

These new members of the VAX family—combined with ever-improving

networking capabilities—provided significant and varied configuration
possibilities for DIGITAL customers.

25
CHAPTER VII Networking

DIGITAL recognized early on that its customers needed a means of

connecting various systems and coordinating their capabilities. To address
this need, the company began research in this area as early as 1972, when
it developed a multiprocessor that would combine a number of minicom-
puters to obtain the power of a large mainframe. This was accomplished
through networking.

By 1973, DIGITAL formed a group to direct the design and implementation

of a networking project; its goal was to achieve absolute compatibility and
interconnectivity across all computer families. The company’s initial efforts
resulted in the DIGITAL Network Architecture that implemented a layered
protocol approach. This method of connecting systems was recognized as
state-of-the-art technology and put DIGITAL in the leadership position in
the industry.

DECnet
In 1974, DIGITAL introduced
DECnet, the industry’s first general-
purpose networking product for
distributed computing. One goal
was to make networks affordable
so that customers could implement
them more widely.

DECnet for VAX and VMS V1.0 was

available for the first customer ship-
ment of the VMS operating system,
and significantly contributed to the
success of the VAX-11/780 system.

Unlike earlier networking

products—which focused on con-
necting terminals to hosts—DECnet
provided peer-to-peer networking
for the first time. This was a major
step toward the client/server
computing model. DIGITAL had developed the best and least expensive
distributed computing solution and became an industry leader with this
technology.

DECnet linked DIGITAL systems together in a flexible network that could

be adapted to changing requirements. It provided direct communication
among computers at the same organizational level, and had no hierarchical
requirements or prerequisite host processors. DIGITAL networks were
modular and flexible—as opposed to IBM’s rigid, hierarchical products—
and could connect computers from other vendors, providing a degree of
compatibility among different computer systems that was unmatched in
the industry.

26
 Over the years, DECnet evolved through five releases, each designed to
Five phases of DECnet work with the next and previous phase. DIGITAL also contributed to the
Phase I: Supported point-to-point major networking standards, incorporating key standards such as OSI and
(directly wired connections) and task- TCP/IP into DECnet.
to-task (customer applications could
be coded to talk to each other over Enter Ethernet
the networking protocols). Ethernet communications capabilities were incorporated into DECnet
Phase II: Added remote file access Phase IV, allowing DECnet users to extend their networks with local area
and general task access (i.e., an capabilities of Ethernet.
application could invoke general
command procedures on a remote The era of the Ethernet brought an entirely new concept to networking.
system). This version of DECnet was DIGITAL set the standards with Xerox and Intel by establishing Ethernet
supported by VMS V1.0, thus VMS as the industry choice for local area networks. The three companies jointly
had remote file access built into the defined the Ethernet standard, which led to the deployment of local area
base file system from networks. Ethernet became the medium-speed but long-distance network,
day one. connecting components as far apart as a kilometer.

Phase III: Added routing, which CI, NI, and BI interconnects

meant you no longer had to have DIGITAL coined the terms CI, NI, and BI as part of an effort to rationalize
a directly wired connection between the company’s strategy for interconnecting the components of computer
two systems to allow them to interact systems at different levels of implementation.
via DECnet. Rather, network traffic
could be forwarded between two sys- NI—Network Interconnect. This was the highest level interconnect,
tems by one or more connecting computer systems in a network. NI quickly became synony-
intervening routing nodes. It also pro- mous with Ethernet. Ethernet allowed the construction of local area
vided SET HOST (the ability to log networks of up to a thousand connections and a mile and half in size.
into a remote system interactively)
and MAIL—the beginnings of corpo- CI—Cluster Interconnect. The CI also connected individual computer sys-
rate electronic mail. tems. In contrast to the NI, it allowed much smaller configurations: up to
Phase IV: Added Ethernet support. 16 systems spread over a 90-foot radius. What the CI lacked in scale it
Ethernet eliminated the requirement made up for in speed—allowing communications over 10 times as fast as
for point-to-point wiring, allowing the NI. DIGITAL developed storage controllers that connected to the CI,
many systems to be connected to a providing the basis for clustered VMS systems (see next section).
single wire in a Local Area Network.
Phase IV also provided a larger BI—Backplane Interconnect. The backplane was used to connect compo-
address and the concept of areas nents of a computer system within a single cabinet. The BI was built to be a
(analogous to telephone area codes), faster replacement for the UNIBUS, used by all PDP-11s and the initial VAX
thus allowing a network to grow to as systems. The VAX 8200 and 8300 used the BI as their “native” interconnect
large as 65,000 nodes. (i.e., both I/O and main memory). Later VAXes (other 8000 and 6000 series)
used the BI strictly to connect to I/O controllers.
Phase V: Incorporated OSI standard
networking into DECnet. Supported SI—Storage Interconnect. A standardized connection between a storage
unlimited address space/nodes device (disk or tape) and its controller.
when using OSI addressing;
supported 100,000 nodes if XI—Everything Interconnect. A future interconnect that would replace NI
using large local files. and CI, being both faster and larger than either. Something like the XI was
ultimately realized with FDDI, but displaced neither the CI nor NI.

The next logical step of networking was computer clustering—a concept

that DIGITAL pioneered. Today, the company continues its position as the
industry leader in clustering.

27
CHAPTER VIII The Second VAX Generation

The VAX 8600

In October 1984, DIGITAL announced the VAX 8600. This system marked
the beginning of the second generation of VAX machines—and a new
milestone in the VAX strategy. The VAX 8600 offered up to 4.2 times
the performance of the VAX-11/780 and increased I/O capability while
maintaining I/O subsystem compatibility with the VAX-11/780 and the
VAX-11/785 Synchronous Backplane Interconnect (SBI).

It was the first VAX implementation in ECL (Emitted Coupled Logic) tech-
nology and the first to include macropipelining. The VAX 8600 represented
the confluence of many new concepts and further refined the solid engi-
neering of earlier systems. It was packaged with an extensive portfolio of
VMS software products that could run on the VAX 8600, as well on all the
earlier models.

The VAX 8600 team.

One platform, one operating system, one network

While DIGITAL had considered as many as eight different approaches to networking,
the company crystallized its approach to networking in 1983 and announced its net-
working strategy at DECworld ‘83. That strategy was one platform (VAX), one operat-
ing system (VMS), and one networking product (Ethernet).

28
VAX production line and test station.

Happy 10th Birthday,

VAX and VMS
The 10th Anniversary of the VAX plat-
form and VMS operating
system in 1987 was celebrated A new high end: the VAX 8800
at DECUS with a VAX-at-10 Dinner In January 1986, DIGITAL introduced its top-of-the-line VAX 8800 and
Speech. The company discussed VAX the midrange VAX 8300 and VAX 8200. These VAX systems were the first
architecture goals and presented an VAX systems to support dual processors. Each machine incorporated a new
overview of the development of VMS. high-performance I/O bus, the VAXBI. The high-performance VAX 8800
achieved application throughput that was two to three times faster than
DIGITAL noted VAX architecture had the VAX 8600.
achieved one of its initial goals—that
of providing a price range span of A year later, the company introduced the VAX 8978 and 8974, the most
1000:1. The company achieved that powerful systems from DIGITAL to date, offering up to 50 times the
goal in February of 1987 power of the VAX-11/780. Both machines included the new 2.5 Gbyte
with the announcement of the SA582 Storage Array from DIGITAL. Combined with the HSC70 I/O
VAXstation 2000, priced at $4,600; processor and the VAXBI bus, the SA482 delivered mainframe-class I/O
while the VAX 8978 was available for subsystem performance array and large storage capacity.
$5,240,000. The company
also discussed how VAX and VMS
grew from FORTRAN-only in 1977
to “101” layered products
in one system in 1987.
“From the late 1970s to the late 1980s, DIGITAL moved from being what I
would call a niche mini-computer company to the second largest computer
company in the world. And that growth was entirely driven by our VAX
and VMS business. From that standpoint, VAX and VMS is one of the really
great success stories in the history of computing, in terms of totally trans-
forming a company and totally transforming an industry and playing a
major role as one of the truly major computer architectures. Certainly VAX
and VMS has been a driving architecture for ten years, and is still a very
important architecture at age 20.”
—Bill Strecker
Chief Technical Officer, VP, CST

29
 Ongoing engineering challenge: Evolving the architecture
By all standards, the VAX and VMS architecture was very stable from the
late 1970s to the 1980s. The reasons for this stability were two-fold. First,
because the architecture was so extensively engineered, it didn’t require
any architectural changes over that ten-year period of time. Second, it
offered virtually everything a customer might want. The architecture had
been designed for longevity, and it succeeded in that goal.

Over a 10-year time period, the product line broadened from a single
VAX-11/780 to a whole family of products that offered continually
improved performance at a lower cost. The next engineering challenge
was to make successively faster implementations of the architecture
at a lower cost.

Don Harbert and Pauline Nist accept While some development teams were working on the traditional VAX
The PC Week corporate satisfaction award systems, others were developing a new breed of chip-based systems,
for the VAX 6000 from Susan Pasieka for which eventually became the main line of DIGITAL’s products.
the second year in a row. – Summer 1992

Bill Demmer with 2nd generation VAX family members.

30
 “When we announced the new VAX
and showcased it at DECUS, cus-
tomers would come up and ask, “What
is that?” We told them that this is a
new VAX. They were pleasantly
surprised.

Our customers loved their VAX sys-

tems. They just put them in a closet
and forgot about them— they’re that
VAX assembly line. reliable. It’s a tribute to the hardware,
the architecture, and the software,
because it’s bullet-proof. The
Catamount was proof that DIGITAL
continues to support its Installed Base
customers.”
– Ed Yee
Senior VAX Product Manager

VAX 6000

The VAX 6000 was the first volume SMP

VAX. In the first six weeks of production,
there were 500 units shipped. The VAX sys-
tem was the so-called tornado of that time
frame—the market just sucked them up. The shipments
grew from a rate of zero to 6,000 units a year in about five
months, which continued for a couple of years.

31
 VAX 8000

VAX Family

VAX 8600

ALPHA AXP 2100

VAX-11/780

VAX 6230

VAX-11/730

1977-1997 ... VAX 6200

AlphaServer 4100

MicroVAX II & MicroVAX 2000

VAXstation I

VAX 8650

VAX-11/782
 VAXstation 2000

VAX 4000 200/300

VAX-11/750

AlphaServer 8200

and beyond
VAXft 3000

VAX-11/730

VAX-11/785

Bill Demmer and the VAX Family

VAX-11/750 Family

VAX Family
CHAPTER IX Putting the VAX on a Chip

Roots of the DIGITAL semiconductor group

The inception of the DIGITAL semiconductor group occurred in the early
1970s with the development of the LSI-11 for the PDP family of computers.
DIGITAL designed the chip and partnered with other companies for
fabrication. By the late 1970s, technological advances in semiconductors
had made chips more powerful and less costly to produce. It became
clear that semiconductor technology was imperative in order to remain
competitive in the computer industry.

Developing the V-11:

The first VAX chip
In 1981, an advanced development
team explored ways to bootstrap
capabilities in semiconductors and
design a full-scale VAX on a chip.
This project, V-11, was intended to
be a full-scale VAX CPU, implement-
ed with state-of-the-art semiconduc-
tor technology—N Channel or
NMOS. As such, it required four dif-
ferent chips in the implementation.

As the project moved forward, it

became clear that microcomputer
systems were going to be built very
differently from the way the V-11
was being built. Microcomputer
systems were going to be based on
single chip microprocessors aimed
at a dramatically lower price.

DIGITAL addressed the question

of whether the VAX design could be
turned from being a minicomputer
architecture implemented in silicon
into a true microprocessor architec-
ture that could be competitive with
industry microprocessors. The
company decided to do the latter.

VAX quality control inspection of VAX 8600

CPU board.

32
 The V-11 resulted in the VLSI VAX chip, which was shipped in the VAX
VAX systems earn their stripes 8200 and 8300 series systems. The V-11 was replaced by the MicroVAX chip,
VAX systems—due to their perfor- but it provided the design technology, basic architecture, and many of the
mance, network capabilities, and building blocks that made up MicroVAX.
scalability—found their way into many
military and Department of Defense Designing the MicroVAX: The first chip-based VAX
applications. Developers developed The V-11 and the MicroVAX I were developed more or less concurrently.
military/DoD programs for The VAX-11/750 was the first DIGITAL system to be designed with LSI
Command, Control, Communication, semiconductor technology, using gate arrays. After the 750, DIGITAL
and Intelligence (C3I) designed the MicroVAX I—one of the first DIGITAL projects to include
applications. The highly scalable VAX silicon compilers—with the consulting help of Carver Meade, a pioneer in
and OpenVMS architecture per- integrated circuit design. Building on the experience of the MicroVAX I,
formed well in the computer rooms the company soon followed with the more powerful MicroVAX II.
and back lines. But there was a need
to bring the VAX technology closer to While the V-11 was designed as a full VAX implementation, the MicroVAX I
the harsh environment of the battle was designed as a VAX subset. The MicroVAX I system was developed
front. in the company’s Seattle facility, headed by Dave Cutler. Because the
MicroVAX I was a much simpler design than the V-11, and because of the
United Technologies’ Norden
use of the silicon compiler tools, it was completed before the V-11.
Systems, a prime contractor located
in New Hampshire, licensed the VAX
DIGITAL explored the option of having one of the industry’s semiconductor
architecture and developed a milita-
companies produce the chip, but decided to do the work internally because
rized version of the VAX on a chip
of the complexity of the task and the aggressive schedule. This proposal
called the MIL VAX II. The
was considered radical because it placed a great deal of faith in the then
system’s cost was five times that of
fledgling chip organization for both design and manufacturing.
a commercial VAX system, but ran
significantly faster than its
Introducing “the first VAX you can steal”
civilian brother. This system met mili-
The MicroVAX project was launched in July of 1982 and the silicon was
tary environmental testing
finished on February 4th of 1984—just 19 months later. It was an achieve-
standards, including temperature,
ment that was unprecedented in the industry. The entire semiconductor
vibration, shock, salt, fog, dust,
organization rallied around this effort and gave the chip top priority in
explosive atmosphere, and
terms of fabrication and debugging. Thus, they were able to demonstrate
humidity. MIL VAX II was suited for
the chip running VMS by August of 1984, field test the system in late 1984,
database management, command,
and ship it in May of 1985.
control, and intelligence operations
aboard ships and airplanes and
Drastically different from any of the earlier VAX systems, the MicroVAX II
in-land installations.
system was wildly successful. It was the first VAX under $20,000. Commenting
Over the years, other VAX systems on its unprecedented affordability and size, Ken Olsen called it “the first
have been “ruggedized” by many VAX you can steal.”
third-party DoD contractors for use in
less severe military applications.
These systems were used on
shipboard and mobile applications
where they had to withstand the
rigors of shock and vibration. “Bob Supnik had come up with this wonderful scheme to build a VAX on a
chip, which became the MicroVAX II chip, ultimately.”
—Jesse Lipcon
Senior VP, UNIX and OpenVMS Systems Business Unit

33
 Building on the success of the
MicroVAX
The success of the MicroVAX II set
the course of development for the
VAX chip family for the rest of the
1980s. By reducing the VAX CPU to
such a small package and exploiting
semiconductor technology, DIGITAL
was able to continually improve
performance at a dramatic rate.

After the introduction of the

MicroVAX II, the company’s
hardware and software engineers worked together to add back four more
instructions out of the commercial instruction set, and the COBOL design-
ers created a version of the compiler that didn’t require the complex
decimal instructions that had been left out.

The MicroVAX II project would not have been possible if the VMS group
had not vigorously supported the whole concept from the outset. When
MicroVAX was first put together, it was a more drastic departure from the
VAX architecture than the final design—particularly with its simplified form
of memory management. But the original VAX memory management was
reintroduced back to MicroVAX II to make it a machine with complete
functionality.

The MicroVAX II was the system that put the VAX CPU on a chip. With
powerful VAX virtual memory, 32-bit computing power, and software
compatibility across all VAX processors, the MicroVAX II microsystem
provided functionality and flexibility that was unparalleled in the industry.

Skyrocketing sales
DIGITAL was showing the largest volumes of VAX systems sales ever. Up
to that point, a highly successful VAX system sold 2,000 units in its lifetime.
MicroVAX sold 20,000 units in its first year.

The VAXstation 2000

The VAXstation 2000 was a step down in size from the MicroVAX II. Like
the MicroVAX II system, it was built around the MicroVAX II chip. Where
the MicroVAX II was housed in a small, desk-side cabinet and supported
a variety of PDP-11 peripheral devices, the VAXstation 2000 came in a
shoebox-sized cabinet. All the essential functions—CPU, graphics display
controller, disk controller, and two serial ports were integrated on a single
circuit board. Its peripherals were limited to a keyboard, monitor, and
mouse, plus up to two fixed disks, and a floppy disk and tape drive. In
return for those limitations, it delivered near VAX-11/780 performance for
a $5,000 entry price. Customers called it “a MIP on a stick.”

34
In its first year, the VAXstation 2000 sold 60,000 systems. This
demonstrated the principle of elasticity—showing that if you have a
capability and you bring its price down, you enhance its marketability.
Now, with unprecedented affordability, everybody wanted a VAX.

CVAX
The company’s second chip was called CVAX—the C stood for CMOS.
A conversion in technology from the earlier NMOS (N channel, metal
oxide semiconductor) to CMOS (complementary metal oxide semi-
conductor) was due to the market’s relentless climbing power
requirements.

This second-generation VLSI VAX microprocessor offered 2.5 to 3.5

times the power of its predecessor. It was the company’s first internally
manufactured CMOS microprocessor. High performance came from
features such as a macro-instruction pipeline, 1 K Byte onchip
datacache, and a 28 entry onchip translation buffer.

The CVAX chip was also much more complicated than the MicroVAX
chip. The engineers had to develop the CPU/Floating Point functionali-
ty in VLSI and develop separate VLSI chips for Memory Control, the
Q-Bus Interface, and a Support Chip which included the Time of Year
Clock and Serial Line Interfaces. The number and complexity of these
chips added significant challenges to the project.

The CVAX chip was introduced in the MicroVAX 3500 and 3600
systems in September 1978. Another CVAX-based system, the
VAX 6000 platform, was announced April 1988.

Incorporating SMP
The CVAX-based VAX 6000 series was the company’s first venture
into symmetric multiprocessing (SMP).

DIGITAL believed that SMP would require tearing VMS up by the roots
and starting over again. However, DIGITAL engineers found a simpler
approach. The places where VMS did interlocking against interrupts
were located and determined to be the points where VMS had to put in
a more formal lock structure for multiprocessors. A very small team
produced a working prototype of VMS SMP in nine months.

SMP was introduced in VMS version 5.0, announced April 1988.

“The sense I always had was that there were four key technical visionaries at
the beginning of MicroVAX: Dave Cutler, with his creation of the MicroVAX
I system for early software development; Bob Supnik, who headed up
MicroVAX chip development and also wrote the microcode; Jesse Lipcon
who headed up MicroVAX II Server Development; and Dick Hustvedt, who
drove the MicroVMS Software Strategy.”
—Jay Nichols
Computer Special Systems, Manager of Engineering

35
 The VAX 6000 and plug-in power upgrades
Introduced in April of 1988, the VAX 6000 system was the most successful
midrange system in the company’s history, with the fastest time-to-market
and the most units sold.

The most significant attribute of the VAX 6000 was that it introduced the
concept of rapid technology-based upgrades. With previous DIGITAL
systems, it wasn’t possible to increase power simply by replacing processor
boards. The VAX 6000 introduced the concept of plug-and-play. In other
words, as a faster processor became available the customer could unplug
the old processor, plug in the new processor, and the original equipment
would never have to be thrown away. This allowed customers to increase
power as they needed—and protect their investments in hardware and soft-
ware.

Rigel
The CVAX chip was soon followed by the Rigel chip, the company’s third
32-bit microprocessor. DIGITAL engineers considered two options for this
chip. One proposal was to base the Rigel chip on the VAX 8800, which was
the company’s most successful machine. The other proposal was to produce
a more elaborate design that would have required multiple chips and more
coordination, thus involving higher risk but higher performance.

Ultimately, DIGITAL chose to replicate the circuit design of the 8800 CPU
board on a single chip—Rigel.

The Rigel chip was manufactured in 1.5-micron CMOS technology.

Introduced in July 1989, the Rigel chip shipped in the VAX 6400 system
and later, in the VAX 4000 system. Rigel also included the first implementa-
tion of the vector extension of the VAX architecture.

Mariah
In October 1990, DIGITAL introduced the Mariah chip set, which shipped
in the VAX 6500. An improvement on the Rigel chip set, the Mariah chip
set was manufactured in 1.0-micron CMOS technology. The VAX 6500
processor delivered approximately 13 times the power of a VAX-11/780
system, per processor. The VAX 6500 systems implemented a new cache
technique called write-back cache, which reduced CPU-to-memory traffic
on the system bus, allowing multiprocessor systems to operate more efficiently.

NVAX
The NVAX chip was introduced in November of 1991. The company’s fourth
VAX microprocessor, the NVAX chip was implemented in 0.75-micron
CMOS technology and shipped in the VAX 6600. The NVAX incorporated the
pipelined performance of the VAX 9000 and was the fastest CISC chip of its
time—delivering 30 times the CPU speed of the VAX-11/780.

The NVAX chip is the current technology used in VAX systems shipping today.

36
Moving at breakneck speed
Chip development at DIGITAL was remarkably speedy. The timeframe
from MicroVAX to CVAX was about two and a quarter years. From CVAX
to Rigel was less than two years. From Rigel to Mariah was about a year.
Mariah to NVAX was 15 months.

Growing the business through silicon

The VAX chip set launched the company’s product development in a new “Reflecting on the different chip sets,
direction. In the first full fiscal year, the VAX chip business grew into a from MicroVAX II through CVAX,
Rigel and NVAX — the primary
billion-dollar business. Ultimately, it grew to a two to three billion-dollar
focus of architectural energy was
business. processor performance, with the
NVAX architecture pushing
When MicroVAX was introduced, less than 10 percent of the company’s sys- creativity to its limits.”
tems revenue came from products based on microprocessor chips. By 1990,
—Jay Nichols
microprocessor chips were responsible for 90 percent of the systems rev-
Computer Special Systems,
enue. By the early 1990s, the DIGITAL semiconductor group was the
Manager of Engineering
largest and most profitable business in the company.

Major performance increases

Powered by MicroVAX chips, VAX systems increased in performance
from one MIP, to 2.5 MIPS, to 7 MIPS, to 11 MIPS to over 30 MIPS in
five generations of design. The VAX system had established a worldwide
reputation as the fastest, highest-performance machine on the market.

DIGITAL measured the performance of its chips against the competition

from the time MicroVAX was introduced. The company’s goal—to produce
the industry’s fastest microprocessors—was achieved with CVAX, which
was the fastest chip of its time.

VAX 9000 chip manufacturing clean room.

37
CHAPTER X Building the Bridge to Alpha

Prism: VMS on RISC technology

DIGITAL began working on RISC technology in 1986 when Jack Smith,
VP of Operations, tapped Dave Cutler on the shoulder and said, “You will be
RISC Czar for DIGITAL. Organize a program.” The program, code named
Prism, was to develop the company’s RISC machine. Its operating system
would embody the next generation of design principles and have a
compatibility layer for UNIX and VMS.

The team discussed such issues as: Should it be 32 or 64 bits? Should it be

targeted for the commercial or technical market? The proposed implemen-
tation of Prism was an ECL machine. While known for being particularly
power-hungry, ECL was the fastest semiconductor technology available
during the 1970s and ’80s. The VAX 8600, 8800, and 9000 series were built
using ECL. However, with the NVAX chip in 1991, CMOS technology
surpassed ECL’s performance at a much lower power cost.

There were already two other ECL projects underway, the VAX 9000 and
a successor to the VAX 8800. Would these machines be competitive or
overlapping in the marketplace? What would be their comparative
performance? What about cost? Obviously, it made no sense for DIGITAL
to be developing three projects of the same magnitude.

In April of 1988, a group of workstation engineers made a counter proposal

to get DIGITAL into the technical computing market via existing RISC
technology. They started building a RISC workstation that would run
ULTRIX—the company’s port of UNIX—using microprocessors from
a startup company called MIPS. Prism was canceled in favor of using
MIPS technology.

February 1991, DIGITAL announced Alpha,

programming for the 21st century.

38
 The speedy MicroPrism chip
Porting VMS to Alpha Meanwhile, the semiconductor group in Hudson, Massachusetts, was
While the Alpha architecture was working on the MicroPrism chip—a single-chip CMOS implementation of
being designed, the principal piece the Prism architecture. After the Prism program was canceled, the Hudson
of work needing attention was VMS. group was allowed to complete the MicroPrism chip, since it was very
Nancy Kronenberg led the VMS chal- near completion. The small batch of MicroPrism chips produced ran
lenge, which seemed rather formida- successfully at 45 MHz—a speed that was unheard of at that time,
ble. VMS contained more than 10 and that far surpassed the performance of any RISC chip available on
million lines of code—much of it writ- the market.
ten in VAX assembly code. It was
coded to all the features of the VAX The birth of Alpha
instruction The Prism program was significant for DIGITAL because of the legacy
set, and it was unclear how to it left for Alpha—the company’s future 64-bit technology. A small team
separate VMS from VAX. formed in July of 1988 to determine what RISC technology could do for
VMS. First the team asked themselves, “What do we have to do to get VMS
Through careful analysis, Nancy’s
up on RISC?” Then they turned the question around. “If the customers
team discovered that even though
have to go through a transition, how do we get the maximum performance
VMS looked monolithic, it was a well-
and minimize their pain?” That’s when Alpha was born.
structured operating system with a
machine-dependent and a machine-
Alpha was very much the “son of Prism.” The primary changes made to
independent layer. The machine-
produce Alpha were for VMS compatibility. The original Prism design had
dependent layer could be ported and
serious compatibility problems with the VAX and VMS in two areas—
the machine-independent layer
numerical data types and privileged architecture.
would follow. The team invented solu-
tions such as the macro compiler,
The Alpha architecture was built on four premises. First, it had to be a very
which treated
long-lived architecture. Second, it had to deliver the highest performance
VAX macro code as a higher level lan-
for both technical and commercial applications. Third, it had to be very
guage and compiled it to Alpha.
scalable in terms of both implementation size and range of systems sup-
In 1991, the final task—porting VMS ported. And fourth, it had to support customers’ applications and operating
to Alpha—fell to Jean Proulx and her systems, VMS and UNIX. Windows NT had not yet entered the scene.
team who accomplished the porting
challenge brilliantly. VMS was Alpha-
ready!

“We’ve done a lot of work to make sure that moving from VAX to Alpha is
very easy. If a customer doesn’t want to move their entire environment to
Alpha at one time, they don’t have to. We support mixed architecture clusters,
which allows VAX and Alpha to run together in a cluster. They can stay on
VAX as long as they’d like to. We’ll continue to do releases of OpenVMS
Alpha and OpenVMS VAX at the same time.”
—Rich Marcello
Vice President, OpenVMS Systems Software Group

39
 Asking the right questions
The team made decisions about the product by asking questions. “If the
objective is to create a 20-year target, will a 32-bit machine be viable 20
years from now?” The answer, “No.” So it became a 64-bit machine. That
part was easy. “What would it take to drive performance over 20 years via
clock rate improvements, multiple instruction issues, internal organization,
and multi-processing?” The architecture reflects exactly what it takes to
do that.

They looked at the issue of scalability from small to large, and therefore
had a model of what could be the minimum implementation. Research
done on Prism helped the team solve operating system, data flexibility,
and code handling issues. Another critical development issue was the
notion of VAX-to-Alpha binary translation to ensure a smooth migration
for DIGITAL customers who would eventually move to 64-bit computing.

Determining Alpha’s building blocks

The basic building blocks of Alpha were: an archi-
tectural commitment to move to 64 bits with the
highest levels of performance that would preserve
DIGITAL customers’ investments, a matching
commitment on VMS to preserve customers’ oper-
ating environments, and silicon that would stand
the industry on its ear. The design team studied
high-speed implementation techniques discovered
through the MicroPrism project. The team con-
cluded that a chip could be built that would run
two to three times faster than anything else in the
industry—one that would run at 200 MHz when
competitors were talking about 50 MHz.

Bringing the company on board

The Alpha program ran as a loose confederation of
people who shared the vision of putting DIGITAL
Ken Olsen visits manufacturing facility during back on top with leadership systems. There was an Alpha project in VMS,
the power-up of the first Alpha system.
and an Alpha project in DIGITAL semiconductor group. These team
members went out and proselytized to the rest of the company and
convinced it group by group to participate, until eventually the Alpha
program consumed roughly a third of the company’s engineering
resources.

40
Getting business partners on board
In order for the partners at DIGITAL to take advantage of this record-
breaking technology, Vice President Bill Demmer set up the Alpha AXP
Partners Office six months before the announcement so that the company’s
business partners would be signed up and on board at announcement time.
Early Alpha partners included Andersen Consulting, Cray Research,
Encore, Kubota Pacific Computer, Raytheon, and Olivetti.

By September 1992, DIGITAL had shipped more than 1,000 Alpha systems
to software developers.

Bringing on the customers

To meet customer needs, DIGITAL developed programs and services to
support this new technology. For two years prior to the 64-bit announce-
ment, a group of customers met regularly to review plans for the Alpha
AXP program. The group, the ALPHA AXP Customer End User Advisory,
included representatives from communications, manufacturing, tech-
nology, government, the university community, and other potential
markets for the Alpha technology.

With peak execution rates of up to 2 BIPS,

these top-performing Alpha 21164 chips push
the performance envelope for visual comput-
ing applications such as video conferencing,
3-D modeling, video editing, multimedia
authoring, image rendering, and animation.

41
Chapter XI AlphaChip—The 64-bit Breakthrough

On February 25, 1992, DIGITAL introduced another significant technology

advance: the world’s first 64-bit architecture. This revolutionary architec-
ture was based on the AlphaChip 64-bit RISC technology and 150 MHz
DECchip 21064 microprocessor.

Announcing the Alpha AXP family of systems

In November 1992, DIGITAL announced a complete family of ALPHA AXP
systems. It included ALPHA AXP workstations, departmental servers, data
center servers, mainframe-class servers, and system software, as well as
services, layered products, peripherals, and upgrade programs. Four
hundred software partners announced availability dates for nearly 900
Alpha applications.

Alpha AXP achieved record-breaking status. In April, ALPHA AXP per-

formed the world’s fastest sort and fastest transaction processing to date.
The company announced the industry’s highest-performance workstations
in the less than $5,000, $10,000, and $15,000 price categories.

Above to the left; Bill Demmer, VP, discusses AlphaChip in 1992 announcement.
Above; Ken Olsen visits Alpha manufacturing facility.

“Today is the beginning of a new revolution in computing. With nearly limit-

less 64-bit computing power and the applications of three major operating
systems, the path ahead leads wherever the imagination can take it. ALPHA
AXP computing will enable customers to invest in profitable new ways to
serve people.”
—Robert B. Palmer
Chairman, President and CEO of Digital Equipment Corporation
Q2 FY93

42
 VAX—enjoying life after Alpha
Many people thought that after
DIGITAL announced its family of
64-bit Alpha computers, there would
be no more VAX systems introduced.
Not so.
In 1995, DIGITAL announced the
Catamount project, which was Ken Olsen and Bob Palmer discuss future
responsible for producing the VAX technology.
4000 Model 108 system and
MicroVAX 3100 Model 88 and 98
systems. DIGITAL added new func-
tionality into the product set, includ-
ing an increased memory capacity by
a factor of four. Engineers increased
the memory capacity in response to
customer requests for more memory
to meet prior increases in CPU per-
formance. The Catamount products
were designed to be both rack Destination Alpha: Removing the barriers
mountable and used on the desktop. To help ensure that customers have a risk-free transition from VAX systems
The focus was lowering the cus- to Alpha systems, DIGITAL launched the Destination Alpha program in
tomers’ cost of ownership and allow- 1995. Under this program, DIGITAL opened 34 application migration
ing customers to take advantage of centers around the globe to help customers migrate their applications.
lower cost memory and storage tech- In addition, an engineering hotline is available to help customers resolve
nology. their most critical migration issues.
Beyond the product enhancements,
the real significance of this new line DIGITAL also developed a program called Project Navigator that addresses
of VAX systems was the fact that DIG- any financial or technical barriers that customers may face. Through these
ITAL was continuing to make invest- programs and services, DIGITAL has provided customers with a smooth
ments to support transition to the Alpha platform.
the company’s Installed Base of VAX
customers.

“When we were designing the Destination Alpha Program, we realized that

we needed to develop customized solutions so customers could move from
VAX to Alpha at their own pace.”
—Janet Darden
Destination Alpha Program Manager

43
 VMS becomes OpenVMS
The 64-bit Alpha system became the most powerful system in the industry.
Major developments included strategic Alpha software combined with the
availability of Microsoft’s Windows NT on the Alpha platform. During this
time frame, DIGITAL officially changed the name of VMS to OpenVMS to
reflect the ease of portability and openness of this operating system. With
OpenVMS, VMS now supported the widely accepted POSIX standards of
the IEEE. The VAX operating system was also “branded” by X/Open, the non-
profit consortium of many of the world’s major information system suppliers.

OpenVMS supports key standards such as OSF/Motif, POSIX, XPG4, and

the OSF Distributed Computing Environment (DCE). Extensive support
for standards in the operating system helps when building an open systems
environment using OpenVMS as the base. Supported open systems stan-
dards include networking, data, document, systems, software development,
and user interface. OpenVMS supports all major open systems standards,
including those for networking, data, document, systems, software
development, and user interface.

VMS becomes OpenVMS.

44
With this name change came the introduction of 13 Alpha-ready OpenVMS
VAX systems and servers. Alpha-ready was the term coined to indicate that
these VAX machines were easily upgraded to incorporate the new 64-bit
technology.

In February 1993, the company shipped 26 OpenVMS ALPHA AXP

products ahead of schedule to provide a software suite for developers,
system integrators, and end users. In May, more than 2,000 applications
were available for OpenVMS ALPHA AXP.

ALPHA AXP family members.

45
CHAPTER XII Inaugurating the King of Clusters

Throughout the industry, increasing demands were placed on computers

as customers’ applications grew. One way to provide more computing
power was to build bigger, faster systems up to the current technical limits.
DIGITAL came up with an alternate solution that provided more power—
without sacrificing the benefits of distributed computing customers wanted.
That ideal was clustering.

Cluster computing, invented by DIGITAL, has become a widely accepted

alternative method of providing higher system availability and scalability
using mainstream computing products than can be provided by a single
computer system. In fact, in the eyes of our customers, DIGITAL’s OpenVMS
Clusters became the standard by which all other clusters are measured.
Cluster computing provides a dimension of scalability as an alternative to
extending or upgrading a single system, and allows older installed systems
to be coupled into the cluster to provide an economical way to increase
computing power and deliver higher availability of data and applications.

Introducing VAXclusters
In May 1983, DIGITAL announced VAXclusters. VAXclusters tied
VAX processors together in a loose processor coupling that allowed
VAX computers to operate as a single system—extending VAX
characteristics to high-capacity and high-availability applications.

OpenVMS Clusters
Over the years, VAXclusters evolved to VMSclusters, and today are
OpenVMS Clusters for VAX and Alpha Systems. OpenVMS Clusters
are unparalleled in the industry today. Most of the world’s stock
exchanges and electronic funds transfer activities run on OpenVMS
Clusters.

An OpenVMS Cluster is a highly integrated organization of VAX and

Alpha systems, application and systems software, and storage
devices. Systems sized from the desktop to the datacenter can be
connected into an OpenVMS Cluster. OpenVMS Cluster software
enables the system to work an easy-to-manage virtual system that
shares printing resources, storage devices, and print and batch
queues.

OpenVMS Clusters offer the best benefits of both centralized and dis-
tributed systems with the added benefit of power that can surpass
that of a mainframe—at a fraction of the cost. And they can be added
to or divided as customer requirements dictate.

VAXclusters was the first clustering

capability in the industry! VAXclusters
tied VAX processors together, which allowed
VAX computers to operate as a single system, “(Open)VMS remains King of the Clusters. DIGITAL’s technology
extending the characteristics of VAX to
high-availability applications. is still the high bar against which other clustering schemes
are measured.”
—Datamation, August 15, 1995

46
 Local Area VAXclusters
In 1986, DIGITAL introduced Local Area VAXclusters, which extended
distributed computing capability to the workgroup and used a standard
Ethernet network as the cluster interconnect.
Unparalleled benefits of
OpenVMS Clusters With Local Area VAXclusters, VMS extended its cluster technology to the
High availability—Guaranteed access NI. The CI interface was a large, expensive controller available only on
to data and applications due to mul- large, expensive VAX systems. That fact, plus the limit at that time of 16
tiple connected systems. systems on a CI, limited CI clusters to the large “computer room” VAX
systems. Also, all cluster-accessible storage had to be connected directly
Easy growth—A cluster can contain
to the CI. However, the advent of the MicroVAX and VAX workstations
anywhere from 2 to 96 systems,
(concurrent with clusters in 1984) created the demand to connect larger
depending on the changing needs of
numbers of smaller VMS systems into the cluster.
the business.
Shared access—All users can To answer this demand, DIGITAL modified VMS to allow the cluster
easily access applications, storage communication protocols to operate over the NI, which was the only
devices, and printers within a interconnect available on small VAX systems. In addition, software was
cluster. introduced to allow all storage devices on the cluster to be served to all
cluster members. This allowed the NI cluster members access to the
Easy to manage—An entire cluster
HSC-based storage even though they had no direct connection.
can be managed as a single system,
remotely or on-site.
Investment protection—Existing
systems can be integrated into the
same cluster along with new VAX and
Alpha technology.
Multiple interconnects—Clusters can
be configured using many different
interconnects, including CI, DSSI,
SCSI, NI, and FDDI.
Automatic caching—Enhances perfor-
mance and reduces I/O activity.
DECamds—Optional availability
management tool allows monitoring
and managing resources availability
in real-time.
Lock manager services—Allows
reliable access to any resource or
file, without the danger of losing
or corrupting the file and its data.

Local area VAXCluster systems extended VAXCluster

technology to Ethernet. Bringing the software advantages
of the VAXCluster environment to the MicroVAX II and
VAXstation II systems.

47
 Supporting clusters via more interconnects
In the years since, DIGITAL added more interconnects to support cluster
connections:

• FDDI—an industry-standard, optical fiber interconnect approximately

ten times faster than an Ethernet. The FDDI also provided access to
bridges to a number of common carrier communications media, allowing
Bring on more mules cluster connections over great distances.
An analogy can be drawn from farm-
ing. To do more plowing, the farmer • DSSI—a low-cost CI that allows connection of up to three VMS systems
can work his mule longer and harder, and a limited number of directly attached disks.
or trade in his old mule for a bigger
and stronger one. Another option is • Memory Channel—a very fast direct memory access path between
to buy a second mule and team VMS systems located close together.
them together.

Clustering, or joining computers

together to share a task, was like
hooking up a second mule. And a
third. And a fourth. Customers could
keep their existing investments and
grow from there.

“The high-availability characteristics of OpenVMS and clusters are very

important to us...the scalability and clustering capability of OpenVMS allow
us to provide our clients with technology as they need it.”
—Scott Fancher
Vice President and Product Line Executive, Cerner Corporation

48
Fault Tolerant and Disaster Tolerant Systems
Just as clustering was an outgrowth of networking, fault tolerant and
disaster tolerant systems were an outgrowth of clustering.

Clusters offer high availability; they are not fault tolerant. Clustering
enabled the development of fault tolerant and disaster tolerant systems by
providing availability that guaranteed 24x365 days of service. Fault tolerant
systems provide what is considered five nines of availability, meaning that
the system would be available 99.999 percent of the time. Fault tolerant
systems allow applications to continue computing in the event of equip-
ment failure. The system does not have to wait to restart or boot after
encountering a failure. Rather, the failed piece of equipment drops off
while the redundant pair continues to run from the time of the fault. In
certain situations this kind of availability is needed—such as 9-1-1 emer-
gency services, financial/stock market transactions, air traffic control, and
nuclear reactor monitoring. Fault tolerant applications are needed where
the consequences are disastrous if the computer is out for a few minutes or
more. Fault tolerant failover occurs in a minute or less, with no loss of data.

Multi-site clustered systems are used in disaster tolerant applications.

Disaster tolerant systems are set up to prepare for man-made or environ-
mental disasters including terrorism, fires, earthquakes, floods, etc. All
these situations have the potential to take out a computer room. If there is
a back-up system that can send out data to another location, the systems
will remain functioning to prevent losses of data and business that an
interruption would cause. Two fault tolerant systems clustered together in
two different locations provide site diversity and automatic failover to a site
distanced from the disaster. If one site goes down, the other takes over and
continues operating—without missing a beat. VAXft 3000 Announcement

OpenVMS Clusters continue to reign as the King of Clusters

Today, more than 65,000 OpenVMS Cluster
systems are found at the heart of continuous
computing solutions for such critical applica-
tions as stock exchanges, electronic funds
transfers, healthcare, telecommunications, and
process manufacturing. No other solution can
match OpenVMS Cluster systems when it comes
to our over 14 years of providing a continuous
computing environment. Only OpenVMS Cluster
systems can span up to 500 miles to enable
continuous operation through even large scale
natural or man-made disasters ensuring optimal
data and transaction integrity and fast recovery.
OpenVMS Cluster system support “rolling
upgrades”, enabling system processors, boards,
peripherals, operating software, databases, and The Fault Tolerant Group
program modules to be replaced, upgrades, or
updated without interrupting the operation.

49
CHAPTER XIII OpenVMS Today

OpenVMS is a general purpose, multi-user operating system that runs in

The OpenVMS Ambassadors both production and development environments. OpenVMS Alpha supports
Program, formerly known as the DIGITAL Alpha series of computers, while OpenVMS VAX supports the
OpenVMS Partners, is an internation- VAX series of computers. The software supports industry standards for
al program that provides a facilitating application portability and interoperability. It also supports
liaison between customers and symmetrical multiprocessing (SMP) support for multiprocessing Alpha
the company’s OpenVMS Systems and VAX systems.
Software Group and expert field
organizations in sales support, An integral part of three-tier computing
systems integration, Technical Today, the core of the OpenVMS strategy is to leverage the inherent affinity
Consulting Center (TCC), and bench- between Windows NT and OpenVMS by combining the unequaled strengths
marking. The OpenVMS of OpenVMS with the emerging power and application library of Windows
Ambassadors provide valuable NT in a seamless computing environment.
customer feedback, and because of
their technical expertise can relay OpenVMS is the environment of choice in the most demanding of continu-
information in engineering terms and ous computing situations. The high levels of availability, integrity, security,
can make recommendations about and scalability of OpenVMS make it a natural unlimited high-end for
what types of changes are needed Windows NT in a three-tier client/server environment. OpenVMS is the
from the customers’ perspective. The number one operating system in healthcare today. It also enjoys a major
Ambassadors must meet three presence in the financial, funds transfer, and stock exchange industries,
essential criteria: technical compe- as well as manufacturing, education, and government.
tence, commitment, and a high level
of
communication skills.

OpenVMS Ambassador Team (Business Partners)

“The OpenVMS operating system environment holds a special place in the computer industry. It was the centerpiece
of the minicomputer revolution, the first operating system to prove that scaling from desktop to data center was
practical, and the first to demonstrate that clustered systems could achieve levels of availability well beyond
mainframes or ‘fault tolerant’ systems. It was, and continues to be, a huge market success.”
—Wes Melling
VP of Windows NT and OpenVMS Systems Group

50
 “Marketing for OpenVMS is really
fun activity. We have the most loyal,
most enthusiastic customer groups
Unparalleled availability Enhanced support for clustering out there. They appreciate the tech-
OpenVMS provides immunity to OpenVMS Cluster technology nology. They appreciate the ease of
planned and unplanned downtime enables customers to configure use. They appreciate the value of an
with proven 24x365 availability, disaster-tolerant multi-site clusters operating system that has become
including disaster-tolerant multi-site located up to 500 miles tried and true over a number of
clusters spanning 500 miles. (800 kilometers) apart. years and has evolved to the state
where many of the world’s largest
OpenVMS systems scale to meet the OpenVMS provides features specifi- banks, stock exchanges, healthcare
performance, availability, and data cally designed to improve perfor- organizations, and production man-
requirements of the largest enter- mance and expand OpenVMS ufacturing environments are trust-
prise applications through 64-bit, Cluster configuration flexibility. ing their business to the true 24x365
Very Large Memory (VLM), and OpenVMS supports mixed architec- capabilities of OpenVMS.”
Very Large Data Base (VLDB) sup- ture clusters and allows customers —Mary Ellen Fortier
port, and clusters of up to 96 nodes. to connect up to 96 Alpha and VAX Director, OpenVMS Marketing
systems and storage controllers to
OpenVMS provides enhanced per- share common data and resources
formance, clustering flexibility, easy across systems, as well as architec-
Internet connection, and 64-bit VLM tures. OpenVMS Cluster systems Supporting 64-bit environments
for business-critical applications. can utilize FDDI, CI, DSSI, In November 1995, at DECUS,
New features have been incorporat- Ethernet, and Mixed-interconnect DIGITAL announced OpenVMS
ed to further performance in transports. Version 7.0—supporting 64-bit
OpenVMS Clustering and to virtual addressing. 64-bits of
improve system management. Two powerful features of OpenVMS address space is 18 exabytes. That’s
Memory Channel clusters, extended Clusters are Memory Channel four billion times the 32-bit address
VLM capability, cluster failover, and and the Business Recovery Server. space of four billion bytes. Using
the OpenVMS Internet Product Suite Memory Channel comprises a 64-bit addressing allows developers
are also provided by OpenVMS. high-performance interconnect to map large amounts of data into
technology for PCI-based Alpha memory to provide high levels of
systems that improves OpenVMS performance and to support very
Cluster performance and reduces large memory systems.
costs. Business Recovery Server
Cluster support allows businesses The current Alpha memory man-
to withstand disasters—floods, agement architecture allows actual
fires, earthquakes—at any site, address space usage of eight ter-
without loss of access to data abytes. On the VAX, only half the
“It is important to emphasize the address space is available for appli-
significance of the Installed Base to or applications.
cations (2GB), so the currently
DIGITAL. With over 700,000 systems available Alpha address space is
installed worldwide, it is more criti- OpenVMS Cluster systems can be
managed centrally, as a single sys- 4,000 times that on the VAX.
cal than ever before for us to contin-
ue to nurture the base. tem, providing a single domain for
data, users, queues, and security. As ever larger memory becomes
We brought a bright future to our available, the Alpha memory man-
OpenVMS Installed Base customers agement architecture can be extend-
with the Affinity strategy. OpenVMS ed to support more of the theoretical
continues to be one of three strategic maximum of 18 exabytes. This was
platforms from DIGITAL.” the largest incremental release in
—Wally Cole the OpenVMS operating system
VP, Installed Base Marketing since the introduction of
VMSclusters.

51
CHAPTER XIV Serving Customers Worldwide

Supporting the family–global services

DIGITAL realized that a key factor in the success of VAX and VMS was
customer services. Almost from its founding, the company has supported
customers worldwide from strategically located field service facilities.

Educational services provide software and hardware training—providing

Bringing in the voice of DIGITAL customers with the necessary skills to implement and work with
the customer the company’s system effectively.
From its inception, DIGITAL has
believed that two-way customer com- DIGITAL developed its Services group to ensure that the first release of
munication was necessary to ensure VMS could be supported by field support services. This group formulated
that the company was building prod- strategies for support of VMS, and learned the software in depth to be able
ucts to solve real-world needs. That to support it and train people in the unique features of the new software. A
strategy exists today, as DIGITAL back-up support group called VAXworks was also formed to address cus-
takes a comprehensive approach to tomer needs. The VAXworks group received phone calls and telexes from
working with customers at all levels people all over the world.
of their organizations.
DIGITAL set out to have the best support and field service operations as
DIGITAL listens to customers through well as the best education and training organization. These services have
a variety of forums, always been a vital part of the company’s success and have contributed
including: greatly to the business.

Customer visits—DIGITAL makes more

than 500 visits to OpenVMS
customers annually.

Technical Direction Forums—

Twice a year, DIGITAL presents
new strategies and technologies
to 12 top customers at the Director of
MIS level. This feedback has a direct
impact on future directions.

OpenVMS Executive Counsel—

Every six months, DIGITAL meets with
35-40 CIOs in various customer orga-
nizations to look at
overall strategies of business Field service engineer repairing customer
and direction. CPU board.

DECUS—Founded in 1961, the Digital

Equipment Computer Users Society
(DECUS) is an opportunity for people
who work with OpenVMS on a day-to-
day basis to receive training on all
technologies and
provide valuable feedback.
“A big part of the success at DIGITAL was the support and the service. We
gave enormous service to the customer. And without that, even VAX and
VMS wouldn’t have been so successful.”
—Ken Olsen, 1997

52
Service strategy today
The range of DIGITAL Service spans the spectrum from systems integra- “We find that customers are using
tion to hardware and software maintenance. The company’s service effort information technology to get
focuses on three areas. The first area is to support the company’s strategic greater access to their data. They
growth areas: high-performance 64-bit computing, NT across the enter- want to spend more time on the
prise, and Internet connectivity. The second area is multivendor service. analysis of this information and its
DIGITAL is the only major vendor that has declared a vendor neutral distribution using the Internet to
strategy. The third area is value-added services and innovation create competitive advantage. They
in the marketplace. don’t want to spend a lot of time
becoming information technology
DIGITAL has an investment in global resources and infrastructure experts. More and more they’re rely-
that’s second to none in the industry. The worldwide DIGITAL Services ing on key service partners to take
Organization—between its Multivendor Customer Services and Systems responsibility for the management of
Integration Organization—includes more than 25,000 Service Professionals this information infrastructure.”
worldwide and over 450 locations around the world. At the company’s —John Rando
Solution Centers, System Integration Specialists and Network Consultants VP and General Manager,
help customers successfully solve their most challenging information Multivendor Customer Services Organization
technology problems.

Strategic partnerships
To ensure continued growth and to meet the changing business needs of its
customers, DIGITAL has established strategic partnerships with industry-
leading companies such as Microsoft Corporation, Oracle Corporation,
and others.

53
 Getting the bugs out. Or, using
cockroaches as semiconductors “When we were about up to Version 3 of VMS, I was at DECUS and a
A VAX-11/780 installed at the Carling customer came up to me and she said “You won’t believe this, but we’re
Brewery was crashing still running Baselevel 5 and we love it. We think it’s the best thing ever.
several times a day with no pattern and we’re never going to change it because it does just what we need.”
at all. Field services reps had So I said, “Well, if it does just what you need, I think you’re right, don’t
replaced everything and they ever change it.”
couldn’t figure it out. Every time —Kathy Morse
the machine crashed, accidents VMS Engineer
would happen, usually spilling large
quantities of beer.
One day, a software specialist was in
the booth with the machine
pouring over the last dump. All VAX: Built to last
of a sudden there was the familiar 780 drops off a forklift and lives
rhythm of another crash. He looked In 1978, a VAX-11/780 was shipped to the National Computer Conference in
out the windows and saw people Anaheim. At the loading dock it dropped off of a forklift—which was hard on some-
scurrying for cover. The capping thing this big. A replacement machine from a nearby local office was brought to the
machine had run amuck and was show, and the carcass of the dropped system was shipped back to New England.
spitting out bottle cap blanks, which The engineers took it apart, straightened the frame, and replaced the backplane.
in their raw state are like little two- Other than that, it worked perfectly, and was put in service for years, many years. It
inch diameter, razor-sharp, aluminum was called the Phoenix.
frisbees. The software specialist
couldn’t take it anymore. He walked Another VAX-11/780 slams into the side of a building and keeps on ticking
over to the VAX-11/780 and kicked Another VAX-11/780 was shipped to a customer in Washington, D.C. It was too big
the front panel as hard as he could. for the elevator, so the customer decided to lift it on a crane and swing it
A bunch of cockroaches came scurry- in through a window. Instead of going into the open window, the system slammed
ing out. into the side of a building. The machine looked very damaged.

Naturally, cockroaches are attracted At that time, VAX-11/780 systems were on a six-month backlog and the
to beer dregs. And it was warm and customers didn’t want to wait for a new one. So the feild service engineers put new
dry inside the machine, so they skins on it and replaced a slightly bent backplane, offering a replacement if neces-
moved in. He figured that cockroach- sary. That machine always worked perfectly. And the customer was delighted
es are at least somewhat conductive. because he got a six-month lead.
As the insects ran up and down the
backplane, every once and a while
one of them would
get two legs across—a pair of
contacts. And the machine would
crash.
The software specialist went out to
the local store and bought Roach
Motels which installed in the bottom “The customers were members of the family, and there was a strong dialogue
of the machine. The problems ended. at all levels between engineers and customers. We spent a lot of time hanging
After that, changing the Roach around listening to customers, and DECUS was very active and effective as
Motels became a part of a lobbying committee for new product requirements. We built what the
the monthly product maintenance. customers told us they needed.”
—Larry Portner
VP of Software Engineering

54
“The VAX-11/780 has always held
a lot of sentimental value for me.
Like your first love, you never for- DECUS—Digital Equipment
get your first computer. An VAX- Computer Users Society
11/780 was the first computer I DECUS, the Digital Equipment
programmed on in the early ’80s. Computer Users Society, is an associ-
My very first program, a test of a ation of Information Technology pro-
FORTRAN subroutine, I jokingly fessionals interested in the products,
called ‘[Link]’ so that I could services, and technologies of Digital
‘RUN FAST’ under VMS. Equipment Corporation and related
Noteworthy because it was the vendors. The Association’s purpose is
first in the very popular VAX line to promote the unimpeded exchange
of of information, with the goal of help-
computers from DIGITAL, the Top; Mastermind of the VAXbar, Vance ing its members and their organiza-
VAX-11/780 is also one of the few Haemmerle. Bottom; Old VAX-11/780 systems tions to be more successful. The
computers that can actually be never die, they just find new ways to serve Chapter provides members with the
considered a classic. Not only did humanity.
means to enhance their professional
it play a pivotal role in the mini- development,
computer revolution, but it also forums for technical training, mecha-
evolved into a standard. nisms for obtaining up-to-date
information, advocacy programs, and
So when a VAX-11/780 donated to opportunities for informal discussion
the HACKS computer club was on the and interaction with professional col-
loading dock headed for the dump leagues of like interest.
after its CPU and memory boards
were stripped for parts to get two
other 780s working, I persuaded fel-
low HACKS members to let me have it
“It was extremely exciting in the early days of marketing VAX and VMS
and find another use for it. My ideas
because the company was growing so fast. I remember the first DECUS that
were either a bookcase for the ‘Grey
I attended. Each time I went to DECUS, the audience doubled. The first time,
Wall’ of VMS books, or a wetbar.
there were about 300 people in the audience, the next time there were 600,
Since I didn’t have my own copy of
the next time there were 1,200, and the next time there were over 2,000. That
the VMS
was the momentum that we had in the market. In many cases we would
documentation at that time, the
learn from our customers about different applications using 32-bit.”
wetbar idea was the obvious choice!
—Vance Haemmerle
—Marion Dancy
VP of Marketing, UNIX and OpenVMS,
Systems Business Unit

55
CHAPTER XV The Affinity Program

Leveraging the natural affinity

of OpenVMS and Windows NT
On May 8, 1995, at DECUS in
Washington, D.C., Digital Equipment
Corporation and Microsoft
Corporation announced the Affinity
for OpenVMS Program to help
customers implement the comple-
mentary strengths of OpenVMS
and Windows NT in a three-tier
client/server environment.
OpenVMS provides the ultimate
high-end, tools, and applications to
ensure a seamless integration with
Windows NT.
Bill Gates and Robert Palmer.
This integrated systems environ-
ment brings the bulletproof capabili-
ties of OpenVMS to the world of
Windows NT applications. The
program includes new software,
tools, middleware, and services
from DIGITAL and its partners that
build on the natural affinity between
OpenVMS and Windows NT—
making it increasingly easier to
develop, deploy, and manage
applications across both platforms.

Since May of 1995, DIGITAL has consistently announced new products,

“When our customers had beta capabilities, features, and services that support the OpenVMS Affinity
copies of Windows NT, they told us
environment.
that it felt like they were revisiting
an old friend. That’s not surprising
because the chief architect of both Key examples include OpenVMS V7.0 for 64-bit computing, new products
operating systems was Dave Cutler. for system management, World Wide Web hosting, enterprise messaging,
So there is a natural affinity from a and application development. In addition, software vendors have responded
technical perspective between the two to user demand with new applications and tools. Each year, more DIGITAL
environments. Wes Melling is often business partners are bringing application development, data warehousing,
quoted calling it the ‘Cutler effect.’” and healthcare applications to the Affinity portfolio.
—Mary Ellen Fortier
Director, OpenVMS Marketing In the two years since its inception, the Affinity for OpenVMS Program has
helped more than 20,000 organizations around the world integrate the two
platforms in three-tier client/server environments across their enterprises.
Customers include worldwide banks and stock exchanges, healthcare
providers, manufacturing facilities, educational institutions, government
organizations, and more.

“The magnitude of what we are doing here—the clustering agreement, joint

engineering, joint field teams—is much bigger than what we have done in
the past with other alliances.”
—Bill Gates
President and CEO, Microsoft Corporation
56
“With the DIGITAL Affinity for OpenVMS Program, OpenVMS and
Windows NT integration brings Corning the richest operating environments.
It provides superior overall functionality and the most complete set of appli-
cations and tools. We’ve selected Forté—the industry’s premier three-tier
client/server software—which draws on the strengths of both OpenVMS and
Windows NT, making it possible to create and deploy applications with
multi-tier, enterprise-wide functionality.”
—Mark Joyce
Supervisor of Fiber Systems Engineering
Computer and Information Services, Corning, Inc.

Credit Lyonnais
“In any disaster, the key is to protect the data. If you lose your CPUs, you
can replace them. If you lose your network, you can rebuild it. If you lose
your data, you are down for several months. In the capital markets, that
means you are dead. During the fire at our headquarters, the DIGITAL VMS
Clusters were very effective at protecting the data ... What impressed us was
the ability of all of our major suppliers to mobilize and furnish equipment
and services. DIGITAL managed this very well indeed. They were everywhere
with us.”
—Patrick Hummel
IT Director Capital Markets Division, Credit Lyonnais

57
CHAPTER XVI Vision of the Future

Looking skyward: Galaxy

DIGITAL knows that a company’s need for computing resources can
fluctuate significantly for certain applications at certain times.

“Today, OpenVMS is the most flexi- For example, let’s consider a scenario of a system manager for a large
ble and adaptable operating system cluster in a telecommunications company. Once every three months, a
on the planet. What started out as communication satellite might send enormous quantities of vital data to his
the concept of ‘Starlet’ in 1975 is receiving station. Transmission time is only two hours, and it’s critical that
moving into Galaxy’ for the 21st
all the data are processed immediately. He gets no second chances. But
century. And like the universe, there
is no end in sight.” his systems are already busy crunching day-to-day information. Short of
buying, or leasing, new CPUs, memory, and disks, what can he do?
—Jesse Lipcon,
Senior VP, UNIX and OpenVMS
Systems Business Unit
That’s where Galaxy will come in. DIGITAL is developing an evolution in
OpenVMS functionality that will include a new model of computing that
allows multiple instances of OpenVMS to execute cooperatively in a single
computer. For companies looking to improve their ability to manage unpre-
dictable, variable, or growing IT workloads, the DIGITAL Galaxy software
solution for OpenVMS provides the most flexible way to dynamically recon-
figure and manage system resources. Galaxy is a powerful software solu-
tion that allows system managers to easily reallocate individual CPUs or
memory through a simple drag-and-drop procedure.

Enhancing OpenVMS
After 20 years, OpenVMS still has tremendous growth potential.
OpenVMS is a key component of the DIGITAL strategy to satisfy
its customers’ computing needs well into the next century.

“The growing importance of the Internet and corporate

intranets perpetuates the value of OpenVMS. This is an area
where 24x365 is essential. DIGITAL offers a variety of Web-based
servers. OpenVMS is a platform that provides full reliability and
availability of Internet services.”
—Harry Copperman
Senior VP and General Manager Products Division

Alpha System, the path to the 21st Century.

58
The five-pronged OpenVMS strategy
“OpenVMS plays a critical role in our customers’ operations. It’s a very
1. DIGITAL will maintain all current vibrant, vital operating system, with exceptional performance and high
OpenVMS capabilities and will ease availability.”
the migration to a 64-bit environ-
—Bruce Claflin
ment. Senior VP and General Manager,
2. The company will continue to invest Sales and Marketing
in OpenVMS development to
ensure the long-term future of the
operating system.
3. The disaster tolerant, 24X365
strengths of OpenVMS will continue
to be enhanced.
4. The company will provide seamless
integration with Windows NT.
5. OpenVMS will continue to provide
an unlimited high-end to Windows
NT. Current OpenVMS engineering
projects are upholding the same
high standards of engineering
excellence that have characterized
OpenVMS from its inception.

DIGITAL will continue to focus on 64-

bit computing in such areas as the
Internet, continuous computing,
Windows NT integration, and data
warehousing. To meet the demand for
the integration of enterprise comput-
ing with Windows NT, DIGITAL will
develop enterprise applications, visu-
al computing, and mail and messag-
ing that are NT-integrated. Internet
business growth will support the mar-
ketplace need for the development of
customer intranets, Internet com-
merce, and ISP/Telco support.

“If OpenVMS engineering continues the type of innovation we’re doing now,
we’ll be here for another 20 years, and then we’ll be asking ourselves again:
what’s next?”
—Steve Zalewski
Technical Director of OpenVMS systems software group

59
VMS TO OPENVMS: Major Releases

VMS V1 August 1978 VAX V4 September 1984 ~ VMS V5.4 October 1990
• Multiuser, multifunction 40,000 licenses • Support for new processors –
virtual memory operating system • Support for new processor – VAX 8600 VAX 6000-510,520
• ODS-1 and ODS-2 file systems MicroVAX I/II (v4.1) • Vector processing option for
• Integrated DECnet VAXstation I/II (v4.1) VAX 6000-4xx
• ANSI magtape support • VAXclusters • DCL commands for Fault Tolerant
• Languages • Connection manager (VAXft) systems
• VAX-11 FORTRAN IV-PLUS • Distributed lock manager • TPU enhancements
• VAX-11 MACRO generates native code • Distributed file system (F11BXQP) • DECwindows enhancements
• BASIC-PLUS 2 and COBOL • Security enhancements • MSCP load balancing and preferred path
• DCL and MCR command language • Command line editing and • Password history. Dictionary and site
interpreters command recall specific password filters
• Supported hardware • Local area terminal server • Hardware releases
• VAX-11/780 with a minimum of 256 KB • Access control lists implemented • V5.4-0A October 1990
of memory, up to a maximum of 2 MB • Cluster wide operator control • VAX 9000
• 2 RK06 disks, or MASSBUS disk • Variable prompt strings • V5.4-1, December 1990, replaced
and tape • VAX 9000 SMP
• DMC-11 communications interface VMS V4.4 • VAXstation 3100 Model 76
• CR11, LP11, and LA11 • Support for new processors – VAX 8200, • VAX 4000 Model 200
• DZ11 with VT52 and LA36 terminals VAX 8250, VAX 8300, VAX 8350 • VAXft models 110, 310, 410, 610, 612
• Floating point accelerator VAX 8500, VAX 8550, VAX 8700,
VAX 8800 VMS V5.5 November 1991
VMS V2 April 1980 ~3000 licenses • ASMP support for VAX 83xx and VAX • Support for new processors – MicroVAX
• Support for new processor – VAX-11/750 88xx systems 3100 Models 30, 40 & 80
• More native languages • Cluster packages VAX 8974 & VAX 8978 VAX & VAXserver 6000-6xx series
• EDT screen editor • Disk volume shadowing and VAX 4000 Models 60, 500 & 600
• SET HOST HSC support VAXstation 4000 Model 60 & VLC
• MAIL, PATCH and SEARCH utilities • New queue manager
• Shared sequential RMS files VMS/ V5 May 1988 • New licensing features
• Support for multiport shared memory • Support for new processors – VAX 6210, • LAT enhancements (SET HOST/LAT,
and DR780 6220, 6230, 6240, 8810, 8820, 8830, 8840, LATmasterfeatures)
• Connect-to-interrupt driver 8842, VAXserver 6210, 6220 • Phase II Shadowing (host based
• User written system services • Symmetric multiprocessing shadowing)
• VAX FORTRAN (77) (SMP) support • Cluster wide tape service (TMSCP)
• Mixed interconnect VAXclusters • New RTLs - DECthreads and BLAS
VMS V3 April 1982 ~10,000 licenses • License management facility fast-vector maths library
• Support for new processors – • Terminal fallback utility • Hardware releases
VAX-11/730, VAX 11/725, VAX-11/782 • Modularized executive • V5.5-2HW September 1992
• Asymmetric multiprocessing (ASMP) • Structured DCL: IF-THEN-ELSE, • MicroVAX 3100 Model 90,
for VAX-11/782 GOSUB and CALL • VAX 4000 Models 100 & 400
• Support for new architectures, protocols, • System Management enhancements • VAX 7000 Model 600
busses • Dynamic failover of dual pathed disks • VAX 10000 Model 600
• System communication architecture • New batch and print queue features • VAXstation 4000 Model 90
(SCS) • AUTOGEN Feedback • V5.5-2 September 1992
• Mass storage control protocol (MSCP) • VAX 7000 Models 610, 620, 630, 640,
• Lock management system services DEC Windows (v5.1) VMS V5.2 800 through 860
• MONITOR utility for September 1989, ~300,000 licenses • VAXstation Model 90A
performance monitoring • Support for new processors – VAX & • V5.5-2H4 August 1993
• BACKUP VAXserver 6400 series, VAXserver 3100 • MicroVAX 3100 Models 85,88,95,96
• Command definition utility for DCL • Clusters of 96 nodes • VAX 4000 Models 100A, 105A, 106A, 108,
• Terminal autobaud detection, CTRL/T, • Hardware release 500A, 505A, 600A, 700A, 705A, 800A
and hangup on logout • V5.2-1 October 1989 • VAXstation 4000 Model 96
• SPAWN and ATTACH • MicroVAX 3100 • V5.5-2HF August 1993
• VAXstation 3100 Model 38/48 • VAXft Model 810
• VAXstation 6000 Series 4XX

60
OpenVMS/AXP V1.0 November 1992 - OpenVMS/VAX V6.2 May 1995 & OpenVMS/VAX V7.1 & OpenVMS/Alpha
Alpha is here! OpenVMS/Alpha V6.2 June 1995 V7.1 December 1996
• Support for new Alpha processors • Support for new processors – • Support for new processors –
DEC 3000 Models 400, 400S, 500 & 500S AlphaServer 2100 5/250, 8200 5/300, AlphaServer 800 5/333 & 5/400
DEC 4000 Model 600 8400 5/300 • Pipes
DEC 7000 Model 610 • Freeware V1.0 CD distributed with • Windows NT Affinity
• Based on VMS V5.4 operating system • PPP protocol
• DECmigrate for translating VAX images • Automatic foreign commands • Internet product suite
• MACRO-32 compiler (like UNIX PATH mechanism) • Dump Off System Disk for Alpha
• No clusters, no RMS journaling, no • RAID subsystem support • External Authentication (LAN manager
shadowing, no SMP • DCL TCP/IP functions e.g.: COPY/FTP single signon)
and SMTP transport in MAIL • 100BaseT Fast ethernet support (Alpha)
OpenVMS/VAX V6.0 June 1993 • OpenVMS Management Station • Memory channel high performance
• Support for new processors – VAX 7000 • SCSI clusters cluster interconnect
Model 650/660, VAX 10000 Model • SCSI-2 Tagged Command Queuing • Very Large Memory (VLM) support
650/660 • BACKUP Manager - Screen oriented • BACKUP API
• Rationalized and Enhanced security interface • CDE interface for DECwindows
(Level C2 compliance) • Hardware releases • 64 bit system services
• Multiple queue managers across cluster • V6.2-1H1 (Alpha) November 1994 • Scheduling system services
• HELP/MESSAGE utility • AlphaServer 1000A 4/266
• Support for ISO 9660 CD-ROM format • AlphaBook 1
• Adaptive Pool Management • AlphaServer 2100A 4/275, 5/250 & 5/300 Compiled by John Gillings Sydney CSC,
• SYSMAN cluster wide SHUTDOWN • AlphaStation 255/233 & 255/300 September 1997.
and startup logging • V6.2-1H2 (Alpha) January 1995 Sources: Ruth Goldenberg, Max Burnet,
• Cluster wide Virtual I/O cache • AlphaServer 300 4/266 Steve Tolna, Thomas Schwarz, Mark
• Extended physical and virtual • AlphaServer 1000A 5/266, 5/333 & 5/400 Buda, Sharon Rogenmoser, Kim Kinney,
addressing • AlphaServer 4000 5/300E Ken Blaylock, Rod Barela, Kelly Oglesby,
• Protected subsystems • AlphaServer 4100 5/400, 5/300, 5/400 & Marie Teixeira, Michael Junge, Julian
• DECnet/OSI 5/466 Sandoval, Mark Masias, Jason Gallant,
• DECwindows XUI replaced by • AlphaServer 8200 5/440 Brian Breton, Laura Buckley, Richard
DECwindows Motif • AlphaServer 8400 5/440 Rhodes, Dave Pina, Sue Clavin, Tim
• AlphaStation 500/300, 500/400 & 500/500 Ellison, John Manning, Dave Hutchins,
OpenVMS/VAX V6.1 April 1994 & • AlphaStation 600 5/266, 600 5/300 & Paul McGrath, Judy Novey, Ian Ring, Ron
OpenVMS/Alpha V6.1 May 1994 5/333 Decker, Stephen Hoffman, VMS marketing,
• VAX and Alpha
Sales Updates, Old PID material, VMS
• Support for new processors – OpenVMS/VAX V7.0 & information sheet ED-31080-48, VMS
AlphaServer 2100 4/200 & 4/275 OpenVMS/Alpha V7.0 December 1995 SPD’s, OpenVMS New Features Manuals
DEC 3000 Models 700 & 900 • Process affinities and capabilities from
DEC 7000 Model 710 & 7xxx DCL (Set PROCESS/AFFINITY) Edited by: Andy Goldstein
VAX 7000 Models 7xxx • HYPERSORT High performance SORT
• PCSI Product installation utility utility (Alpha)
(PRODUCT command) • Ingegrated network and internet support
• Shadowing and RMS Journaling • New MAIL utility (rewritten)
for Alpha • Timezone and UTC support
• DECamds bundled with operating • 64-bit addressing – new system services
system • Kernel threads
• CLUE Crash dump utility • Spiralog high performance file system
• DPML standard maths library • Dump file compression (Alpha)
• C++ support • Wind/U – Windows Win32 API
• DECnet/OSI Extended Node Names • Fast I/O and Fast Path highly
optimized I/O

61
 VAX and VMS History
1975
• VAX architecture committee 1981
formed with goal of “build- • VAX information architecture
ing a computer that is cul- introduced, which included
turally compatible with VAX-11, FMS, DATATRIEVE,
PDP-11, but with increased CDD, RMS, and DBMS.
address space of 32-bits.”
The result: VAX, the “Virtual
Address eXtension” of the
PDP-11’s 16-bit architecture.
• VMS, the “Virtual Memory
System” operating system
was developed simultane-
ously, allowing complete
integration of hardware
and software.

1979 1983
• DECnet Phase II • DIGITAL announced VAXclusters:
announced. the capability of tying VAX processors
together in a loose processor coupling
• Fortran IV announced.
that allowed multiple VAX systems to
1977 operate as a single system.
• Introduction of VAX 11-/780, • VAX-11/725 announced.
the first VAX system. • CI connectivity introduced.
• VMS V1.0 announced.

1980
• VMS V2.0 shipped, offering
the industry’s largest array of
1978 languages on one system.
• VMS V1.0 shipped. The development • DECnet Phase III announced.
goal was to achieve compatibility
• VAX-11/750 introduced, the
between PDP-11 and VAX systems
second VAX family member
so information and programs could
and the industry’s first Large
be shared.
Scale Integration (LSI) 32-bit
1984
minicomputer.
• VAX-11/785 introduced, the
most powerful single VAX
computer to date. CPU cycle
time was 133ns, 50% faster
than the 200ns cycle time
of the VAX-11/780.
• VMS V4.0 announced.
• VAX 8600 announced, the
first of a new generation of
1982 VAX systems. Offered up to
• VAX- 11/730 announced, the 4.2 times the performance
third VAX family member, of the VAX-11/780;
the first low-cost VAX increased I/O capability
processor to fit on 3 hex while maintaining I/O
boards, the first VAX to subsystem compatibility
fit into a 10.5-inch-high with the VAX-11/780 and
rackmountable box. VAX-11/785 systems.
• VMS V3.0 shipped. • VAXstation I announced,
• RA60 and RA81 disk DIGITAL’s first 32-bit
drives shipped. single-user workstation.
 1991
• OpenVMS name change announced.
• NVAX, DIGITAL’s fourth VAX microprocessor,
implemented in 0.75-micrometer CMOS
technology; shipped in VAX 6600 systems.
• OpenVMS V5.5 shipped.
• DIGITAL and Microsoft Corporation
announced alliance allowing Microsoft
Windows to retrieve and exchange data with
1987 local area network servers running DIGITAL
• VAX 8978 and 8974 systems introduced, PATHWORKS software.
DIGITAL’s most powerful systems to date, • DECnet Phase V announced.
offering up to 50 times the power of the
VAX-11/780 system.
• VAXstation 2000 announced, the first
workstation costing less than $5000,
1985 which ultimately became the highest
• MicroVAX chip introduced for volume workstation in the industry.
the MicroVAX II, DIGITAL’s 1989
• New generation of MicroVAX computers • Introduction of the VAX 6500 System,
first 32-bit microprocessor.
unveiled: the MicroVAX 3500 and 3600. DIGITAL’s most powerful and expand-
First chip manufactured with
internally developed semicon- • CVAX chip introduced, the second-gener- able VAX system in a single cabinet.
ductor technology. “VAX-on-a- ation VLSI VAX microprocessor, offering • VMS V5.1 and V5.2 shipped.
chip” had the highest level of 2.5 times the power of its predecessor. • Rigel chip set introduced. Shipped
functionality of any 32-bit The company’s first internally manufac- in VAX 6400 system and later in
processor in the industry. tured CMOS microprocessor. VAX 4000 system.
• VMS V4.2 shipped.

1986 1990
• Top-of-the-line VAX 8800, midrange VAX • DIGITAL announced VAXft 3000
8300, and VAX 8200 announced, the first system. The first fault-tolerant system
VAX systems to support dual processors. in the industry to run a mainstream
Each machine incorporated VAXBI, a operating system (VMS); first system
new high-performance bus. in which every component, including
1988 the backplane, was mirrored.
• VMS V4.5 shipped. • VAX 6000 System platform announced. Built on 3 key
technologies: the DIGITAL CMOS VLSI VAX processor • VAX 6500 shipped with the Mariah
• Local Area VAXclusters systems intro-
(CVAX chip), a symmetric multiprocessing hardware chip set. The processor delivered
duced, extending distributed computing
and software environment, and the VAXB1 I/O approximately 13 times the power
to the Workgroup via the Ethernet and
interconnect. of a VAX-11/780 system, per processor.
bringing the software advantages of the
VAXcluster environment in MicroVAX II • VMS V5.0 shipped in concert with the VAX 6200 • VMS V5.4 shipped.
systems. system.
 1995
• Affinity for OpenVMS and Windows NT
program announced.
• Affinity Wave I announced – Application
1993 Vendor Partnering.
• OpenVMS AXP V1.5 shipped; • OpenVMS Alpha V6.2 shipped;
OpenVMS VAX 6.0 shipped. OpenVMS VAX V6.2 shipped. 1997
• DIGITAL 2100 Alpha AXP • Wave IV announced – Future
• VAX4000 Model 106A and VAXstation 4000
Server announced. strategy for unlimited high-end
Model 96 announced.
• OpenVMS VAX V7.1 shipped;
• Turbolaser AS8400/AS8200, AS 400 announced.
OpenVMS Alpha V7.1 shipped.
• MicroVAX 3100 Model 96 announced.
• AlphaServer 800 announced.
• AlphaServer 1200 announced.

1994
• OpenVMS VAX V6.1 shipped;
OpenVMS Alpha V6.1 shipped.
• VAX4000 Model 505A/705A.
announced.
1992 • MicroVAX 3100 Model 85
• DIGITAL announced Alpha, 64-bit
announced.
processor architecture for 21st century
computing, Engineered to support
1996
• Affinity Waves II and III
multiple operating systems and
announced.
designed to increase performance by
a factor of 1000 over its 25-year life. • Wave II – Real World
The first Alpha chip was the 21064, Deployment.
which provided record-breaking • Wave III – Advanced Partner
200-Mhz performance. Deployment.
• First-generation Alpha systems • OpenVMS Alpha V7.0 with
included the DEC 3000 Models 400 64-bit VLM/VLDM support
and 500 workstations, DEC 4000 shipped; OpenVMS VAX 7.0
system, DEC 7000 System, and shipped.
DEC 10000 System. • VAX 7000 Model 800, VAX 4000
• MicroVAX 3100 Model 40 announced. Model 108, and MicroVAX 3100
• OpenVMS AXP V1.0 shipped. Model 88 & 98 announced.
• AlphaServer 4000/4100,
AlphaServer 1000A and
AlphaServer 300 announced.
In the very long term, we see
OpenVMS as the huge, bullet-
proof, 24x365, disaster-tolerant
data store for NT applications, in
general. That kind of absolute no-
excuses availability isn’t going to
be matched for a long time by
anybody. In an Internet world,
more and more of our customers
need that availability right now.
—Wes Melling
VP of Windows NT and
OpenVMS Systems Group
The following are trademarks of Digital Equipment Corporation: DIGITAL, International Business Machines Corporation. IEEE and POSIX are registered
the DIGITAL logo, ALPHA, ALPHA AXP, AlphaChip, AlphaServer, BI, Business trademarks of The Institute of Electrical and Electronics Engineers, Inc. Intel is
Recovery Server, CI, DATATRIEVE, DEC 3000, DEC 4000, DEC 7000, DECamds, a registered trademark of Intel Corporation. Microsoft, Windows, and WIN32
DECchip, DECmigrate, DECnet, DECsystem, DECWORLD, FMS, HSC, HSC70, are registered trademarks and Windows NT is a trademark of Microsoft
LSI-11, MicroVAX, MSCP, OpenVMS, PDP, PDP-11, PDP-11/70, RSTS, RSX-11M, Corporation. MIPS is a trademark of MIPS Computer Systems, Inc. Motif and
RT-11, SA, SBI, UNIBUS, ULTRIX, VAX, VAX FORTRAN, VAX 4000, VAX 6000, OSF/Motif are registered trademarks of Open Software Foundation, Inc.
VAX 6400, VAX 6500, VAX 8200, VAX 8300, VAX 8600, VAX 8800, VAX 9000, MUMPS is a registered trademarks of Massachusetts General Hospital. Olivetti
VAXBI, VAXcluster, VAXft, VAXserver, VAXstation, VAX-11/730, VAX-11/750, is a registered trademark of Ing. C. Olivetti. Scrabble is a registered trademark
VAX-11/780, VAX-11/782, VAX-11/785, VLM, VMS, VMScluster. of Milton Bradley. Xerox is a registered trademark of Xerox Corporation.
X/Open is a trademark of X/Open Company Limited. UNIX is a registered trade-
Third-party trademarks: Andersen Consulting is a registered trademark of mark in the United States and other countries, licensed exclusively through
Arthur Andersen & Co. BASIC is a registered trademark of the Trustees of X/Open Company, Ltd. All other products or company names are used for iden-
Dartmouth College, D.B.A. Dartmouth College. Energizer and Energizer Bunny tification purposes only, and may be trademarks of their respective owners.
are registered trademarks of Eveready Battery Company, Inc. Forte is a regis-
tered trademark of Forte Software, Inc. IBM is a registered trademark of

Printed in U.S.A. Rel.# 97/97 10 57 20.0 Copyright 1997 Digital Equipment Corporation. All Rights Reserved.