Compaq ProLiant 5000 Disk Subsystem Performance and Scalability - Page 24
Summary of Findings - Like Capacity Scaling, Disk Controller Scaling
View all Compaq ProLiant 5000 manuals
Add to My Manuals
Save this manual to your list of manuals |
Page 24 highlights
ECG025.0997 WHITE PAPER (cont.) ... Summary of Findings - Like Capacity Scaling Our test results conclude that by doubling the number of drives in a system, regardless of the data storage capacity requirements and the fault tolerance used, we consistently received an improvement in performance. However, the performance increase lessened as we added more and more drives to our system. We should also consider the limitations of each configuration. For example, by using lower capacity drives on one controller you improve performance, however, you sacrifice using the maximum disk capacity on that controller. To decide which test configuration best fits your needs, we recommend that you first review your system requirements, weigh the advantages and disadvantages of like capacity scaling, then purchase the correct drive capacity along with the correct number spindles for your environment. Refer to the "Performance Measurement Tools" section within this document for information on tools that can assist you in measuring system performance. DISK CONTROLLER SCALABILITY When analyzing controller scalability, the focus is on the performance difference between a single controller using a specified number of spindles versus multiple controllers using equally divided drives. Does one controller with all drives attached to it out perform two controllers with the drives divided equally among the controllers? For example, let's say you configure fourteen disk drives in one controller, then split the fourteen drives (seven on each controller) and configure them in a two-controller environment. Which one out performs the other? Disk controller scaling will answer this question for us. Disk Controller Scaling The key to disk controller scaling is to find the point at which your hardware does not produce a significant benefit in system performance. We define this point by testing the scalability of a controller. To scale a controller, we used a constant number of drives with equal disk capacity and tested a single controller versus multiple controllers. For example, the testing results determine if two SMART-2 Controllers with a fewer number of drives on each controller provide better performance than one SMART-2 Controller configured with all the drives. 24