The packet-by-packet switch operates in the identical fashion as a traditional router.

Because the
forwarding logic is in hardware, the packet-by-packet switch can achieve an order of magnitude
increase in performance compared to the software-based router. A flow-based switch tries to
enhance performance by identifying flows of IP packets that have the same source and destination.
This can be done by observing ongoing traffic or by using a special flow label in the packet header
(allowed in IPv6 but not IPv4). Once a flow is identified, a predefined route can be established
through the network to speed up the forwarding process. Again, huge performance increases over a
pure software-based router are achieved.

LLC specifies the mechanisms for addressing stations across the

medium and for controlling the exchange of data between two users.
The operation and format of this standard is based on HDLC. Three
services are provided as alternatives for attached devices using LLC:
• Unacknowledged connectionless service: This service is a
datagram-style service. It is a very simple service that does not
involve any of the flow- and error-control mechanisms. Thus, the
delivery of data is not guaranteed. However, in most devices, there
will be some higher layer of software that deals with reliability issues.
• Connection-mode service: This service is similar to that offered by
HDLC. A logical connection is set up between two users exchanging
data, and flow control and error control are provided.
• Acknowledged connectionless service: This is a cross between the
previous two services. It provides that datagrams are to be
acknowledged, but no prior logical connection is set up.

The basic LLC protocol is modeled after HDLC and has similar
functions and formats. The differences between the two protocols
can be summarized as follows:
• LLC makes use of the asynchronous balanced mode of operation of
HDLC, to support connection-mode LLC service; this is referred to as
type 2 operation. The other HDLC modes are not employed.
• LLC supports an unacknowledged connectionless service using the
unnumbered information PDU; this is known as type 1 operation.
• LLC supports an acknowledged connectionless service by using two
new unnumbered PDUs; this is known as type 3 operation.
• LLC permits multiplexing by the use of LLC service access points
(LSAPs).

All three LLC protocols employ the same PDU format (Stallings DCC8e
Figure 15.7), which consists of four fields. The DSAP (Destination
Service Access Point) and SSAP (Source Service Access Point) fields
each contain a 7-bit address, which specify the destination and
source users of LLC. One bit of the DSAP indicates whether the DSAP
is an individual or group address. One bit of the SSAP indicates
whether the PDU is a command or response PDU. The format of the
LLC control field is identical to that of HDLC (Figure 7.7), using
extended (7-bit) sequence numbers.
For type 1 operation, which supports the unacknowledged
connectionless service, the unnumbered information (UI) PDU is used
to transfer user data. There is no acknowledgment, flow control, or
error control. With type 2 operation, a data link connection is
established between two LLC SAPs prior to data exchange.
Connection establishment is attempted by the type 2 protocol in
response to a request from a user. Once the connection is
established, data are exchanged using information PDUs, as in HDLC.
The information PDUs include send and receive sequence numbers,
for sequencing and flow control. The supervisory PDUs are used, as in
HDLC, for flow control and error control. Either LLC entity can
terminate a logical LLC connection by issuing a disconnect (DISC)
PDU. With type 3 operation, each transmitted PDU is acknowledged.
A new (not found in HDLC) unnumbered PDU, the Acknowledged
Connectionless (AC) Information PDU, is defined. User data are sent
in AC command PDUs and must be acknowledged using an AC
response PDU.

All LANs and MANs consist of collections of devices that must share
the network's transmission capacity. Some means of controlling
access to the transmission medium is needed to provide for an
orderly and efficient use of that capacity. This is the function of a
medium access control (MAC) protocol. The key parameters are
where and how.
Where refers to whether control is exercised in a centralized or
distributed fashion. In a centralized scheme, a controller is
designated that has the authority to grant access to the network. A
station wishing to transmit must wait until it receives permission
from the controller. A centralized scheme has certain advantages,
including greater control over access, the use of relatively simple
access logic at each station, and avoids problems of distributed
coordination among peer entities. The principal disadvantages of
centralized schemes are that it creates a single point of failure, and it
is a bottleneck, reducing performance. In a decentralized network,
the stations collectively perform a medium access control function to
determine dynamically the order in which stations transmit.
The second parameter, how, is constrained by the topology and
is a tradeoff among competing factors, including cost, performance,
and complexity. In general, we can categorize access control
techniques as being either synchronous or asynchronous. With
synchronous techniques, a specific capacity is dedicated to a
connection. Such techniques are generally not optimal in LANs and
MANs because the needs of the stations are unpredictable.
It is preferable to be able to allocate capacity in an asynchronous
(dynamic) fashion, more or less in response to immediate demand.
The asynchronous approach can be further subdivided into three
categories: round robin, reservation, and contention.
With round robin, each station in turn is given the opportunity
to transmit. During that opportunity, the station may decline to
transmit or may transmit subject to a specified upper bound. When it
is finished each station relinquishes its turn, and the right to transmit
passes to the next station in logical sequence. Control of sequence
may be centralized or distributed.
For stream traffic, reservation techniques are well suited. In
general, for these techniques, time on the medium is divided into
slots, much as with synchronous TDM. A station wishing to transmit
reserves future slots for an extended or even an indefinite period.
Again, reservations may be made in a centralized or distributed
fashion.
For bursty traffic, contention techniques are usually
appropriate. With these techniques, no control is exercised to
determine whose turn it is; all stations contend for time in a way that
can be, as we shall see, rather rough and tumble. These techniques
are of necessity distributed in nature. Their principal advantage is
that they are simple to implement and, under light to moderate load,
efficient. For some of these techniques, however, performance tends
to collapse under heavy load.
Although both centralized and distributed reservation
techniques have been implemented in some LAN products, round-
robin and contention techniques are the most common.
The MAC layer receives a block of data from the LLC layer and is
responsible for performing functions related to medium access and
for transmitting the data. As with other protocol layers, MAC
implements these functions making use of a protocol data unit at its
layer. In this case, the PDU is referred to as a MAC frame. The exact
format of the MAC frame differs somewhat for the various MAC
protocols in use. In general, all of the MAC frames have a format
similar to that of Figure 15.7 (see previous slide). The fields of this
frame are
• MAC Control: protocol control info needed for MAC protocol
functioning
• Destination MAC Address: destination physical attachment point
on LAN
• Source MAC Address: source physical attachment point on LAN.
• LLC: LLC data from the next higher layer.
• CRC: The Cyclic Redundancy Check field, an error-detecting code
In the LAN protocol architecture, the functions of detecting
errors using the CRC and retransmitting damaged frames are split
between the MAC and LLC layers. The MAC layer is responsible for
detecting errors and discarding any frames that are in error. The LLC
layer optionally keeps track of which frames have been successfully
received and retransmits unsuccessful frames.