Compaq ProLiant 1000 I/O Performance Tuning of Compaq Servers - Page 19

I/O Performance Tuning of Compaq Servers 19 requested data is retained in high-speed RAM after the request has been filled. If this file is accessed frequently, subsequent requests can be filled immediately, without requiring the hard drive to fetch data. By adding more RAM to your server, you can allocate additional memory and increase the amount of data that can be cached. When properly implemented, cache can enable your server to fill more client requests faster. The increase in performance comes from not only filling the request from faster RAM, but also the fact that the busses between the memory and drives were not used to get the data. In other words, by not using the Fibre Channel or SCSI busses to access this request, these busses are free to handle other requests. In turn, the other requests, which may not be in cache, can be retrieved more quickly. Using larger and better-tuned caches eliminates redundant bus traffic for commonly accessed data. P In general, adding RAM increases I/O, as well as overall, performance. As with other hardware systems, adding more RAM is a case of diminishing returns. That is, the more RAM you have installed, the smaller the gain of adding more will be. Additionally, some operating systems have limits on the amount of RAM that can be used efficiently. Check with the operating system developer to see if your OS is subject to any RAM limitations. Each operating system implements data caching differently. Therefore, to maximize the investment in RAM, make sure that your OS is making the most efficient use of the available RAM. Novell NetWare, Microsoft Windows NT, SCO UnixWare, and SCO OpenServer have many cache tuning parameters that can optimize data caching to your environment and workload. OS Performance Monitoring Tools In every server environment the effect of each bottleneck will differ. Each operating system provides specialized tools for assessing the load placed on your server and the impact each bottleneck has on server performance. The first step in server performance tuning is understanding the sources of performance limitations. The next step is to identify which sources of limitation are most prevalent in your server. Finally, tune hardware and configure the operating system to eliminate or reduce the effect of these bottlenecks. In order to identify which hardware elements may be limiting your server performance, each operating system provides tools to monitor and record hardware performance data. In addition, the monitoring tools provide insight into the operating system itself, providing information about how the OS is using its resources. While many modern operating systems are "auto-tuning," manual adjustment of the server environment variables and registry entries may be necessary to achieve optimal server performance. Most of the OS-variable tuning modifies caching or hardware allocation algorithms. By making accurate adjustments to the caching and OS variables, you can mask or reduce hardware bottlenecks. In this section, we will introduce the OS-specific performance monitoring tools available in Microsoft Windows NT, Novell NetWare, SCO UnixWare, and SCO OpenServer. Each tool provides different insight into server performance. None of these tools provide specific analyses or recommendations on tuning. Rather, these tools gather general information that is not hardware or environment specific. With some intuitive thinking and careful analysis, you can use this insight to identify which bottlenecks are most noticeable in your system. From the initial gathering of data, to the final implementation of changes, performance tuning is not a straightforward process. To effectively use these tools, you should • Collect relevant, useful data. • Be aware of unusual spikes in the data. ECG044.0399