By Evan Morris
In Part I of this series (see InfoStor, July 2002, p. 38), we looked at the process of preparing to stage a storage resource management (SRM) deployment. In this article, we continue the testing phase, provide some sample tests and tools to use in your lab evaluation of SRM tools, and give tips on preparing for the pilot phase rollout and actual deployment.
You may not need to do performance testing or benchmarking of your file systems and servers because the impact of SRM software should be negligible. However, we advise establishing baseline performance so that in the future, you have a frame of reference to determine whether the servers are performing adequately after deploying the SRM software. To test performance impact, we performed before and after snapshots. In our tests, we focused on measuring the performance drag of SRM quota-management software on our file server.
I use several benchmarking tools, including NetBench, Iometer, Ntiogen, Nbench, Bench32, ThreadMark, and HD Tach, to test server perfor mance. (For descriptions of these benchmarks, and links for more information, visit www.wquinn.com/ebook.)
We were already using StorageCentral SRM, so we tested during the upgrade process to version 5.0 because the upgrade requires an uninstall of version 4.1.
The test configuration included:
- Dual-processor Compaq ProLiant 1850 server, 550MHz Pentium III, 512KB cache, 1GB RAM
- Dual Compaq 100BaseT adapters, each to a different network segment
- Win2K SP2 (with security hotfixes)
- SmartArray 4200 RAID controller, 56MB cache, firmware 1.30
- Two disk arrays and three logical disks, all formatted with NTFS - 9GB RAID 1 internal for operating system and pagefile - External array with 14 9GB drives configured for RAID 5, two logical disks
- Up to 16 Compaq Deskpro clients (load generators) using 100BaseT network cards, on the same switched network segment.
The NetBench benchmark tests how well a file server handles file I/O requests from Windows clients and reports throughput and client response time measurements. NetBench includes standard test suite files. In this case, I used the enterprise test suite, ent_dm_nb702.tst, which is designed for larger server configurations.
The figure on the bottom shows the NetBench test results before the SRM software was installed, establishing a baseline for normal file server operations. Establishing a baseline allows you to repeat the tests later to evaluate changes to the system. My server peaked out and leveled off at around 50. The figure on the top shows the NetBench results after we installed the SRM software, running the same test suite.
As shown in the figures, there is a slight difference in file server perfor mance, as expected when adding another layer of software. Is this performance penalty acceptable for the SRM functionality gained? Yes, because the SRM software allows you to turn your back on the servers for extended periods of time and not have to sort through file listings to weed out old and unused files. Also, you don't have to worry about the disks filling up.
In addition, I used the Iometer test to determine the maximum I/O capacity for my disk subsystem before and after I deployed the SRM software. Iometer is an I/O workload generator and performance analysis tool for disk and network stress testing. (For complete test results, visit www.wquinn.com/ebook.)
SRM products usually have more than one quota type, unlike the Win2K quota functionality, which can only apply quotas toward users and groups. In your tests, be sure to include any types of quotas or storage policies that will be applied toward managed objects in production.
Communication and education plan
As part of your pilot program, you should develop and refine communication plans. This step communicates storage policies to the pilot users. You should also develop a user education program.
A FAQ format works well for an intranet site. Questions to be addressed include
- What is storage resource management?
- What is the organizational policy on storage?
- How much storage is available, and where?
- What types of files are allowed?
- Why is this change being made?
- Who can users contact?
- Where can users get more information?
In larger organizations, the pilot phase involves two groups: the technical group of users and the business pilot group. The business group can provide valuable feedback about how the SRM solution either helps or hinders their daily jobs. After performing the lab evaluation, it is time to test the SRM software in a staging of the "live" deployment.
The goal of the pilot is to maximize knowledge about the SRM solution and to minimize problems during the production rollout. In addition to stability testing, you will gain feedback in other areas such as measuring the effectiveness of the implemented solution.
This feedback will come from the system testers and also from the end users involved in the pilot phase.
Preparing for deployment
At this point you should be ready to answer the question: "Are you ready to deploy the SRM solution?" Are you confident that the deployment will go smoothly, with little interruption to users? Remember that software deployment is an ongoing process of learning and revisions, and you will need to revise the solution as needed.
At this point, you should develop an escalation plan and define the production team roles. The template in the table (bottom, left) is an example that can be used for assigning SRM support boundaries and roles.
In this article, we looked at testing SRM software, getting feedback, and assessing the effectiveness of the solution. We focused on project-management tasks, provided sample templates, and assisted in creating a test plan for evaluating the SRM solution. We also showed you how to get started with performance testing and covered the results of our own testing. Finally, we looked at the communication and end-user education plan to convey the appropriate messages about your SRM deployment.
Evan Morris is an MCSE, MCT, Compaq Master ASE, and a messaging consultant in Compaq's Exchange Solutions Engineering division. He writes for Windows 2000 Magazine's Exchange Administrator publication and teaches MCSE certification programs at several universities.
Editor's Note: This article was excerpted from the free eBook, The Definitive Guide to Storage Resource Management (www.realtimepublishers. com), written by Evan Morris and hosted at www.wquinn.com/ebook. Results of all benchmark tests conducted as part of this review, as well as additional templates and instructions, are available at this site.