Performance Tuning for Content Manager

84 Performance Tuning for Content Manager

Periodically, and especially after significant changes to your system or workload:

 Re-examine your objectives and indicators.

 Refine your monitoring plan and tuning methodology.

3.5 General performance tuning guideline

When you are working on improving system performance, we recommend that

you follow this set of general guidelines for performance tuning. This is a

collection of wisdom from the performance tuning experts:

 Establish quantitative, measurable, and realistic objectives.

 Understand and consider the entire system.

 Change one parameter at a time.

 Measure and reconfigure by levels.

 Consider design and redesign.

 Remember the law of diminishing returns.

 Recognize performance tuning limitationss.

 Understand the configuration choices and trade-offs.

 Do not tune just for the sake of tuning.

 Check for hardware as well as software problems.

 Put fall-back procedures in place before you start tuning.

Establish quantitative, measurable, and realistic objectives

Performance is measured in response time and throughput. Performance tuning

objectives must be quantitative, measurable, and realistic. Units of measurement

include one or more of these: response time for a given workload, transactions

per second, I/O operations, CPU use. A system performance can also be

measured in terms of throughput, or how much specific workload such as loading

and archiving can be completed over a specified time.

Understand and consider the entire system

Never tune one parameter or one portion of a system in isolation. Before you

make any adjustments, understand how the entire system works and consider

how your changes could affect the entire system.

Change one parameter at a time

Change one performance tuning parameter at a time to help you assess whether

the changes have been beneficial. If you change multiple parameters at once,

you might have difficulty in evaluating what changes contribute to what results. In

addition, it might be difficult for you to assess the trade-off you have made. Every

time you adjust a parameter to improve one area, there is a great chance that it

affects at least one other area that you might not have considered. By changing

Chapter 3. Performance tuning basics 85

only one at a time, this gives you a benchmark to evaluate whether the result is

what you want to accomplish.

Measure and reconfigure by levels

Tune one level of your system at a time, for the same reasons that you would

change one parameter at a time. Use the following list of levels within a Content

Manager system as a guide:

 Hardware

 Hardware configuration

 Operating system

 Library Server configuration

 Resource Manager configuration

 Database manager

 Database configuration

 WebSphere Application Server

 Application configuration

Consider design and redesign

Design plays an important role in system performance. If the design of a system

is poor, it might be impossible to improve performance and achieve business

requirements. When considering improving a system’s performance, the impact

of the design on a system should be considered. Sometimes, you need to go

back to the drawing board, review the entire system design, and work with

architects and developers to redesign the application or system if necessary to

achieve desired performance. There might uncertainty whether redesign is

considered part of performance tuning; nevertheless, design should be

considered and if necessary redesign should be followed.

Remember the law of diminishing returns

Greatest performance benefits usually come from a few important changes in the

initial effort. Minor changes in the later stage usually produce smaller benefits

and require more effort. Always remember the law of diminishing returns: Stop

when the performance effort no longer justifies the outcome and when the

outcome is already acceptable.

Recognize performance tuning limitations

There are times when tuning improves performance dramatically if a system is

encountering a performance bottleneck, which can be corrected. Other times,

tuning will not make any additional difference because the system is already

performing at its optimal limit. There is a point beyond which tuning can no longer

help. You must recognize this limit. When this situation arises, the best way to

improve performance might be to add additional disks, CPU, or memory or

replace the existing ones with faster and higher performance hardware.