Performance measurement conditions

General

Every performance test case states some conditions that must apply when the performance test is executed.

There are two types of conditions that can be specified:

  1. Static conditions.

  2. Dynamic conditions.

Read more about the importance of recorded dynamic conditions.

1. Static conditions

Static conditions apply equally during an entire performance test.

Examples of static conditions are:

  • The configured hardware platform of the tested system, such as:

    • The number and types of servers.

    • Processor types and number of cores.

    • Memory size and storage type and sizes.

    • Number of communication links, link speeds.

  • The configured software platform of the tested system, such as:

    • The type and version of used operating system.

    • The type and versions of used application middleware, such as databases, web servers, message systems, etc.

Static condition elements in a test bed don't change during a performance test, but may change between two consecutive performance tests.

It's important when comparing test results that static conditions are recorded and verified.

To top of page

2. Dynamic conditions

Dynamic conditions are directly related to the applied work load.

Examples of dynamic conditions are:

  • The actual use of physical resources, such as:

    • Processor load.

    • Memory usage.

    • Disk load.

    • Load on communication links etc.

  • The actual queues for physical resources, such as:

    • Processor queues for CPU resources.

    • Memory swap rates.

    • Disk I/O queues.

    • Communication links queues.

  • The actual queues for application resources, such as:

    • Application threads.

    • Memory buffers.

To top of page

The importance of recorded conditions

The stated dynamic conditions measured during the execution of a test case are key to the quality of processed performance data.

This means that specified dynamic conditions must be recorded in parallel to application performance figures.

A small example: Measure response time for transaction type "A" when CPU load on the tested system is 80%. In this case response time measurements for transaction type "A" is wasted if concurrent measurements of CPU load are not included, since it's impossible to verify the conditions for measured response times.

All performance data recorded when the specified conditions don't apply are invalid and shall be excluded when the measurement results are calculated.

To top of page

Proceed to Load and System Characteristics