Translating resource requirements into performance objectives

Your estimated resource requirements are in important input into the process of defining performance objectives.

Procedure

To translate your resource requirements into performance objectives:

  1. For each workload type, convert your estimated resource requirements into measurable performance objectives.
    Include the following factors when you consider your estimates:
    System response time
    You cannot guarantee requested response times before any of the design has been done. Therefore, plan to review your performance targets when you design and implement the system.

    Response times can vary for many reasons. Therefore, include acceptable tolerances in your descriptions of targets. Remember that distributed data processing adds overhead at both the local and remote locations.

    Exclude from the targets any unusual applications that have exceptionally heavy requirements for processing or database access, or establish individual targets for those applications.

    Network response time
    Responses in the processor are likely to be in microseconds, whereas responses in the network with appropriate facilities can be about a millisecond. This difference in response times means that an overloaded network can impact the delivery of server responses to user terminals or distributed applications regardless of the speed of the processor.
    Disk response time
    I/O operations are generally responsible for much internal processing time. Consider all I/O operations that affect a workload.
    Existing workload.
    Consider the effects of additional work on existing applications. In planning the capacity of the system, consider the total load on each major resource, not just the load for the new application.
    Business factors
    When calculating performance estimates, concentrate on the expected peak throughput rate. Allow for daily peaks (for example, after receipt of mail), weekly peaks (for example, a Monday peak after weekend mail), and seasonal peaks as appropriate to the business. Also allow for peaks of work after planned interruptions, such as preventive maintenance periods and public holidays. Remember that the availability of input data is one of the constraints on throughput.
  2. Include statements about the throughput rates to be supported (including any peak periods) and the internal response time profiles to be achieved.
  3. Make assumptions about I/O rates, paging rates, and workloads.