The advisors analyze the Performance Monitoring Infrastructure
(PMI) data of WebSphere Application Server using general performance principles,
best practices, and WebSphere Application Server-specific rules for tuning.
Procedure
- Ensure that PMI is enabled, which is default. If PMI
is disabled, consult the Enabling PMI using the administrative console topic.
To obtain advice, you must first enable PMI through the administrative console
and restart the server. The Performance and Diagnostic Advisor enables
the appropriate monitoring counter levels for all enabled advice when PMI
is enabled. If specific counters exist that are not wanted, or when disabling
the Performance and Diagnostic Advisor, you might
want to disable PMI or the counters that the Performance and Diagnostic Advisor enabled.
- Click Servers > Application servers in the administrative
console navigation tree.
- Click server_name > Performance and Diagnostic Advisor Configuration.
- Under the Configuration tab, specify the number of processors
on the server. This setting is critical to ensure accurate advice
for the specific configuration of the system.
- Select the Calculation Interval. PMI data is
taken over time and averaged to provide advice. The calculation interval specifies
the length of time over which data is taken for this advice. Therefore, details
within the advice messages display as averages over this interval.
- Select the Maximum Warning Sequence. The maximum
warning sequence refers to the number of consecutive warnings that are issued
before the threshold is updated. For example, if the maximum warning sequence
is set to 3, then the advisor sends only three warnings, to indicate
that the prepared statement cache is overflowing. After three warnings, a
new alert is issued only if the rate of discards exceeds the new threshold
setting.
- Specify Minimum CPU for Working System.
The minimum central processing unit (CPU) for a working system refers
to the CPU level that indicates a application server is under production load.
Or, if you want to tune your application server for peak production loads
that range from 50-90% CPU utilization, set this value to 50. If
the CPU is below this value, some diagnostic and performance advice are still
issued. For example, regardless of the CPU level if you are discarding prepared
statements at a high rate, you are notified.
- Specify CPU Saturated. The CPU
saturated level indicates at what level the CPU is considered fully utilized.
The level determines when concurrency rules no longer increase thread pools
or other resources, even if they are fully utilized.
- Click Apply.
- Click Save.
- Click the Runtime tab.
- Click Restart. Select Restart on the Runtime
tab to reinitialize the Performance and Diagnostic Advisor using
the last configuration information that is saved to disk.
This action also
resets the state of the Performance and Diagnostic Advisor.
For example, the current warning count is reset to zero (0) for each message.
- Simulate a production level load. If you
use the Performance and Diagnostic Advisor in
a test environment, do any other tuning for performance, or simulate a realistic
production load for your application. The application must run this load without
errors. This simulation includes numbers of concurrent users typical of peak
periods, and drives system resources, for example, CPU and memory, to the
levels that are expected in production. The Performance and Diagnostic Advisor provides
advice when CPU utilization exceeds a sufficiently high level only. For a
list of IBM business partners that provide tools to drive this type of load,
see the topic, Performance: Resources for learning in the subsection of Monitoring performance with
third-party tools.
- Select the check box to enable the Performance and Diagnostic Advisor.
Tip: To achieve the best results for performance tuning, enable
the Performance and Diagnostic Advisor when a
stable production-level load is applied.
- Click OK.
- Select Runtime Warnings in the administrative console under
the Runtime Messages in the Status panel or look in the SystemOut.log file,
which is located in the following directory:
profile_root/logs/server_name
Some messages are not issued
immediately.
- Update the product configuration for improved performance, based
on advice. Although the performance advisors attempt to distinguish
between loaded and idle conditions, misleading advice might be issued if the
advisor is enabled while the system is ramping up or down. This result is
especially likely when running short tests. Although the advice helps in most
configurations, there might be situations where the advice hinders performance.
Because of these conditions, advice is not guaranteed. Therefore, test the
environment with the updated configuration to ensure that it functions and
performs better than the previous configuration.
Over time, the advisor
might issue differing advice. The differing advice is due to load fluctuations
and the runtime state. When differing advice is received, you need to look
at all advice and the time period over which it is issued. Advice is taken
during the time that most closely represents the peak production load.
Performance
tuning is an iterative process. After applying advice, simulate a production
load, update the configuration that is based on the advice, and retest for
improved performance. This procedure is continued until optimal performance
is achieved.
What to do next
You can enable and disable advice in the Advice Configuration panel.
Some advice applies only to certain configurations, and can be enabled only
for those configurations. For example, unbounded Object Request Broker (ORB)
service thread pool advice is only relevant when the ORB service thread pool
is unbounded, and can only be enabled when the ORB thread pool is unbounded.
For more information on Advice configuration, see the topic,
Advice configuration settings.