IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Performance White Paper

December 2008

Wesley Varela IBM i Performance Systems and Technology Group International Business Machines Corporation

IBM System Storage DS5300 with seven EXP5000 expansion enclosures in a rack

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Table of Contents
Introduction........................................................................................................................ 3 Test Environment and System Configurations ................................................................. 4 Performance Results .......................................................................................................... 6
CPW - User Series..................................................................................................................... 6

Save and Restore ................................................................................................................ 9

Save and Restore Workloads ................................................................................................... 9 Configuration ............................................................................................................................ 9

Conclusions ...................................................................................................................... 11 Disclaimer System Performance .................................................................................. 12 Notices and Trademarks .................................................................................................. 12 References ........................................................................................................................ 13 DS5000 Disk Storage Subsystem Architecture, Implementation and Usage ................ 13

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Introduction
With more complex customer storage requirements, Storage Area Network (SAN) solutions are becoming the de facto standard. IBM i Performance would like to present the DS5300 Disk System performance measurements. The DS5300 is an excellent Mid to High range storage system that increases performance and capacity relative to the high end DS4000 series by leveraging the latest software and hardware technology. The DS5300 is perfectly suited to leverage the benefits inherent in the 4 Gbps Fiber Channel technology with sixteen 4 Gbps host channels to quickly transfer data, as well as, the 16 GB of physical cache memory (8 GB per controller). The DS5300 can access a total of sixteen EXP5000 or EXP810 expansion drawers for a total of 256 attached disk drives. Typical high throughput and data-intensive applications that would benefit from the DS5300 include (but not limited to) V.O.D (Video On Demand), Medical Imaging, Data mining/warehousing, advanced replication services and, of course, Online Transaction Processing (OLTP). In an Effort to show how the DS5300 compares to an IBM i internal storage solution we have completed several tests using CPW (Commercial Processing Workload) and various Save/Restore workloads using RAID5, RAID6 and RAID10 to demonstrate the particular characteristics of the DS5300 as compared to a completely internal solution. While there are a great many reasons to consider an external storage solution, we highly recommend reviewing several of the documents available online and contacting your IBM Representative to help you find the best external storage solution for your business requirements.

IBM System Storage Product Guide

http://www-03.ibm.com/systems/resources/systems_storage_resource_pguide_prodguidedisk.pdf

IBM System Storage DS5000 series

http://www.ibm.com/common/ssi/fcgibin/ssialias?infotype=PM&subtype=SP&appname=STG_TS_USEN&htmlfid=TSD03061USEN&attachment=TSD03061USEN.PDF

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Test Environment and System Configurations

All measurements where taken using CPW*and Save/Restore. CPW is an internal tool used for Online Transaction Processing (OLTP) measurements. Commercial Processing Workload (CPW). This workload is characterized by many jobs running brief database transactions in an environment that is dominated by IBM system code performing these database operations. CPW is used by IBM for product evaluation of different IBM i models where the primary application is oriented to traditional commercial business uses (order entry, payroll, billing, etc.) using commitment control. Although the CPW workload is typically used to represent the relative performance of a complex of processors, it does a significant number of read and write DASD accesses capable of saturating the I/O subsystem making CPW particularly useful in disk characterization. Save and Restore Workloads. These workloads include: o Large Database File. This is a single database file with one member. It is 64 gigabyte(GB) in size. o User Mix. This is a mix of objects including source files, database files, programs, command objects, data areas, menus, query definitions, etc. It is 12GB in size with approximately 49,500 objects. IBM i Power System configuration used in these tests**: Seven processors of an IBM i 8 way 9117-MMA P6 570. Both Internal and External configurations used the same 8 disk system ASP RAID5 which was hosted in the DASD. 128GB main store where used on both internal and external partitions o Internal Configuration: 96 Disks 72 - Database ASP Mirrored, RAID5 and RAID6 24 - Journal ASP unprotected Six 571B storage adapter cards (90 MB cache per card) 4 for Database ASP, 2 for Journal o DS5300 1818-53A Configuration***: 96 DDMs 12 LUNS (72 DDM) Database ASP Mirrored, RAID5 and RAID6 4 LUNS (24 DDM) Journal ASP unprotected Two 576B 4Gb FC adapter Each port on both cards where connected to different controllers Each line was connected to different cards in the controller. (2) DS5300TM Host Groups: (6) Database LUNS and (2) Journal LUNS comprised each Host group VIOS was running on 1 CPU with 2 GB main store.

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment The DS5300 1818-53A has 16 GB cache (8 GB per controller) 6 EXP810 Enclosures where used. Each LUN was constructed using one disk from each expansion as illustrated in figure 1.

Figure 1

* **

IBM i 6.1 Type 4328 (146GB, 15K RPM) internal disks attached to the storage adapters

During both internal and external evaluations, no other system activity was present. Tests were designed to compare equal number of physical drives not equal capacity. *** External configuration hosted the Journal ASP and the Database ASP. System ASP was on the IBM i Power System. The DS5300TM was configured using Storage Manager V10.30.

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Performance Results
CPW - User Series
Using CPW we are able to test the system configuration using a wide array of methodologies. One such method is a User Series test, in this test a specific number of users are released at discrete intervals. During each interval we can evaluate the I/O subsystem to monitor disk characteristics as a whole, per ASP, per controller even per LUN, as well as over all application response time. For our internal and external configuration tests we started at 16K users and released an additional 16K users at each time interval to a maximum of 176K users with each user attempting to perform 1.22 transactions per minute in each time interval.
R10 Internal Vs DS5300K
0.3 Application Response Time 0.25 0.2 0.15 0.1 0.05 0 0 25000 50000 75000 100000 125000 150000 175000 200000 225000 Transactions Per/Min Internal R10 External R10

Figure2 Application Response Time

As our tests were performed with the CPW user series method, each point on the charted line in Figure 2 represents an additional 16K users being added to the workload. Figure 2 shows the disk response time for internal and external DDM as the number of operations increase to the database ASP. If we examine the External RAID10 (mirrored) line and the Internal RAID10 (mirrored) line we will see a slight increase in the (average) disk response time as the number of operations per second increase, which is not unexpected; management overhead on any external storage solution will be present. However it is important to note that both mirrored runs had an impressive Application Response Time of 1 millisecond at the high end of the workload. This effectively mitigates the trade off between the tremendous flexibility of an external storage solution (FlashCopy, VolumeCopy and Enhanced Remote Mirroring) and the increased latency over internal disks. Figures 3 examine the database ASP and Journal ASP separately.

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

R10 DS5300 vs Internal Database ASP 300 Throughput in MB/Sec 250 200 150 100 50 0 0 16 32 48 64 80 96 11 12 14 16 17 19 2 8 4 0 6 2 Users in thousands DS5K R10 Throughput DS5K R10 Response Time 0.02 0.01 0 0.05
Throughput in MB/Sec 100

R10 DS5300 vs Internal Journal ASP 0.001 Response time in Seconds 0.0008 0.0006 0.0004 0.0002 0 64 80 96 112 128 144 160 176 192 Users in thousands Internal R10 Throughput Internal Response Time DS5K R10 Throughput DS5K R10 Response Time

0.04 0.03

Response time in Seconds

80 60 40 20 0 0 16 32 48

Internal R10 Throughput Internal Response Time

Figure3 Database and Journal ASP Response Times

It is important to point out that Journal LUNS shared a host group connection with Database LUNS, with an increase in response time in a single Host Group can affect but the Journal and the Database ASP response times. Figure 5 examines the RAID5 and RAID6 solutions on the DS5300, while figures 6 and 7 examine the individual ASPs. As you can see, the DS5300TM has advanced caching algorithms and dedicated CPUs that allow it to excel. The DS5300TMs internal cache handily overpowers the internal solutions adaptor cards for the RAID5 and RAID6 workloads.

DS5300 R5, R6 compared to Internal R5, R6

1 0.9 Application Response Time 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 25000 50000 75000 100000 Raw TPM Internal R6 Internal R5 External R5 External R6 125000 150000 175000 200000

Figure 5 R5 and R6 comparisons

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

R5 DS5300 vs Internal Database ASP

160 Throughput in MBytes Sec 140 120 100 80 60 40 20 0 0 0.09 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 16 32 48 64 80 96 112 128 144 160 176 192 Users in thousands Internal R5 Throughput Internal R5 Response Time DS5K R5 Throughput DS5K R5 Response Time

R5 DS5300 vs Internal Journal ASP

100 90 80 70 60 50 40 30 20 10 0 0 16 32 48 64 80 96 0.0016 0.0014 Response time in Seconds 0.0012 0.001 0.0008 0.0006 0.0004 0.0002 0 112 128 144 160 176 192

Throughput in MBytes Sec

0.08 Response time in Seconds

Users in thousands Internal R5 Throughput Internal R5 Response Time DS5K R5 Throughput DS5K R5 Response Time

Figure 6 R5 Database and Journal ASP Views

R6 DS5300 vs Internal Database ASP 160 140 Throughput in MBytes Sec 120 100 80 60 40 20 0 0 0.02 0 16 32 48 64 80 96 112 128 144 160 176 192 Users in thousands Internal R6 Throughput Internal R6 Response Time DS5K R6 Throughput DS5K R6 Response Time 0.08 0.06 0.04 0.12 Response time in Seconds 0.1

R6 DS5300 vs Internal Journal ASP 100 Throughput in MBytes Sec 80 60 0.001 40 20 0 0 16 32 48 64 80 0.0005 0 96 112 128 144 160 176 192 0.002 0.0015 Response time in Seconds

Users in thousands Internal R6 Throughput Internal R6 Response Time DS5K R6 Throughput DS5K R6 Response Time

Figure 7 R6 Database and Journal ASP Views

Recall that CPW is an OLTP (Online Transaction Processing) workload. While latency on individual disk response time increases slightly in a DS5300TM the application response time and transactions per/min that are the hallmarks in this category of testing, were very impressive. Application response time is a weighted average of all transaction response times in a given time interval. Transactions per/min indicate the raw number of accumulated transactions in that interval. Figure 2 through 3 clearly indicate that the DS5300TM performs impressively as compared to internal storage using the CPW workload.

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Save and Restore

Save and Restore Workloads It is important to recall that Save and Restore performance is both data and system specific. Results may vary based on the data, platform and configuration used in the testing. When comparing individual results to the results achieved in this document, it is important to keep these consideration in mind. Note that these save and restore measurements were all done to save files. To determine save and restore rates to other various devices see the IBM Power Systems Performance Capabilities Reference IBM i operating system Version 6.1 in the references section of this document. Chapter 15 on Save Restore Performance focuses on save and restore rates and factors that affect them. Chapter 15.12 specifically addresses virtual tape, which was the implementation in this document. Configuration Measurements were conducted to test the performance of DS5300 external storage RAID 5, RAID 6 and RAID 10 versus internal storage RAID 5, RAID 6 and RAID 10 (mirroring) while running two specific performance save/restore workloads, large database file (single database where all members are at least 4 GB in size) and user mix (single library made up of a combination of source files, database files, programs, command objects, etc). The external or internal disks under test were placed in ASP 2 which contained the save file. ASP 9 contained the workloads. Data was transferred from ASP 2 to ASP 9 for the save and from ASP 9 to ASP2 for the restore. ASP 9 is a virtual tape consisting of 72 type 4328 (146GB, 15K RPM) internal disks attached to (2) 571F I/O Adapters were used to insure any limiting factors would not be related to the configuration of the virtual tape and adequately test the internal and external storage solutions. Here are the results of the save and restore runs:

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Save Performance
Internal RAID 10 External RAID 10 1400 Internal RAID5 External RAID5 Internal R6 External RAID 6

1200

1000 Save Rate (GB/hr)

800

600

400

200

0 Large DB File User Mix

Figure 4 Save performance measurements

The Save portion (figure 4) of the workload shows that the performance is a function of disk response time in ASP 9 as well as the utilization of the 4 GB Fiber Chanel connection between the DS5300TM and our IBM i platform. The external storage solution performed exceptionally well, approaching 1000 GB/hr save rate on all large database file workloads, clearly placing it in the same equivalence class as the internal storage solution. Note that the tests were designed to compare equal physical hardware and not equal capacity between RAID and mirroring, which for mirroring, effectively halves the number of arms, creating the drop in the save rate for the internal mirroring configuration compared to the internal RAID 5 configuration. The Restore portion of the workload effectively leverages the 16 GB of cache to significantly outperform the internal storage solution (figure 5). In the workload that tests the restore of large database files, the DS5300TM almost doubles the restore rate of the internal storage solution and performs equivalently on the user mix files.

10

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Restore Performance

Internal RAID 10 External RAID 10 2500

Internal R5 External RAID 5

Internal R6 External RAID 6

2000 Restore Rate (GB/hr)

1500

1000

500

0 Large DB File User Mix

Figure 5 Restore performance measurements

Conclusions
Incorporating a DS5300TM into any environment has its advantages and disadvantages. The DS5300TM up to offers sixteen 4 Gbps Fiber Channel interfaces for highperformance host connectivity that will deliver high throughput and sustained bandwidth for high performance applications. Additionally, connecting to a Storage Area Network (SAN) over a high speed network can offer tremendous flexibility and reliability. While additional latency exists in the storage management of any SAN product the DS5300TM offers impressive performance results compared to a completely internal solution. Proper planning and sizing considerations are essential in effectively deploying any SAN solution. Please consult Chapter 5 of Sizing external storage for i5/OS in the iSeries and IBM TotalStorage: A Guide to Implementing External Disk on IBM eServer i5 Redbook(s) for more information. Both Redbooks are linked in the reference section of this paper.

11

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment

Disclaimer System Performance

Performance results set forth in this document are based on measurements and projections using standard IBM benchmarks in a controlled environment. The actual throughput or performance that any user will experience may vary depending upon considerations, including, but not limited to, the amount of multiprogramming in the user's job stream, the I/O configuration, the storage configuration, and the workload processed. Therefore, no assurance can be given that an individual user will achieve throughput or performance characteristics or improvements equivalent to the ratios stated here.

Notices and Trademarks

Copyright 2008 by International Business Machines Corporation. No part of this document may be reproduced or transmitted in any form without written permission from IBM Corporation. Product data has been reviewed for accuracy as of the date of initial publication. Product data is subject to change without notice. This information may include technical inaccuracies or typographical errors. IBM may make improvements and/or changes in the product(s) and/or programs(s) at any time without notice. References in this document to IBM products, programs, or services does not imply that IBM intends to make such products, programs or services available in all countries in which IBM operates or does business. THE INFORMATION PROVIDED IN THIS DOCUMENT IS DISTRIBUTED "AS IS" WITHOUT ANY WARRANTY, EITHER EXPRESS OR IMPLIED. IBM EXPRESSLY DISCLAIMS ANY WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. IBM shall have no responsibility to update this information. IBM products are warranted according to the terms and conditions of the agreements (e.g., IBM Customer Agreement, Statement of Limited Warranty, International Program License Agreement, etc.) Under which they are provided. IBM is not responsible for the performance or interoperability of any non-IBM products discussed herein. The performance data contained herein was obtained in a controlled, isolated environment. Actual results that may be obtained in other operating environments may vary significantly. While IBM has reviewed each item for accuracy in a specific situation, there is no guarantee that the same or similar results will be obtained elsewhere. Statements regarding IBMs future direction and intent are subject to change or withdraw without notice, and represent goals and objectives only. The provision of the information contained herein is not intended to, and does not, grant any right or license under any IBM patents or copyrights. Inquiries regarding patent or copyright licenses should be made, in writing, to:

12

IBM System Storage DS5300 Performance Results in IBM i Power Systems Environment IBM Director of Licensing IBM Corporation North Castle Drive Armonk, NY 10504-1785 U.S.A.

IBM, System Storage, IBM i, Power Systems, DS5300 and i5/OS are trademarks of International Business Machines Corporation in the United States, other countries, or both. Other company, products or service names may be trademarks or service marks of others.

References
IBM Power Systems Performance Capabilities Reference IBM i operating system Version 6.1

Performance Management, published October 2008, http://www-03.ibm.com/systems/resources/pcrm_oct2008.pdf DS5000 Disk Storage Subsystem Architecture, Implementation and Usage SG24-7676-00 Draft Redbooks, last update 8 January 2009 http://www.redbooks.ibm.com/redbooks/pdfs/sg245250.pdf IBM System Storage Solutions Handbook, SG24-5250-06 Redbooks, published 12 October 2006, last updated 29 May 2007 http://www.redbooks.ibm.com/redbooks/pdfs/sg245250.pdf iSeries and IBM TotalStorage: A Guide to Implementing External Disk on eServer i5, SG24-7120-00, Redbooks, published 16 October 2005 http://www.redbooks.ibm.com/redbooks/pdfs/sg247120.pdf

13