Os Performance Monitoring Tools - Compaq 108164-003 - ProLiant - 800 Technical Manual

I/O Performance Tuning of Compaq Servers
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
19