Assessing the performance of your system
The following performance measurements can be helpful in determining the performance of a system: processor usage, I/O rates, terminal message or data set record block sizes, paging rates, and error rates.
- Processor usage
- This item reflects how active the processor is. Although the central processor is of primary concern, 37X5 communications controllers and terminal control units can also increase response time if they are heavily used.
- I/O rates
- These rates measure the amount of access to a disk device or data set over a given period. Again, acceptable rates vary depending on the speed of the hardware and response time requirements.
- Terminal message or data set record block sizes
- These factors, when combined with I/O rates, provide information about the current load on the network or DASD subsystem.
- Indications of internal virtual storage limits
- These indications vary by software component, including storage or buffer expansion counts, system messages, and program abends because of system stalls. In CICS®, program fetches on nonresident programs and system short-on-storage or stress messages reflect this condition.
- Paging rates
- CICS can be sensitive to a real storage shortage, and paging rates reflect this shortage. Acceptable paging to DASD rates vary with the speed of the DASD and response time criteria.
- Error rates
- Errors can occur at any point in an online system. If the errors are recoverable, they can go unnoticed, but they put an additional load on the resource on which they are occurring.
Investigate both system conditions and application conditions.
System conditions
A knowledge of the following
conditions can help you evaluate the performance of the system as
a whole:
- System transaction rate (average and peak)
- Internal response time and terminal response time, preferably compared with transaction rate
- Working set, at average and peak transaction rates
- Average number of disk accesses per unit time (total, per channel, and per device)
- Processor usage, compared with transaction rate
- Number of page faults per second, compared with transaction rate and real storage
- Communication line usage (net and actual)
- Average number of active CICS tasks
- Number and duration of outages
Application conditions
Application conditions,
measured both for individual transaction types and for the total system,
give you an estimate of the behavior of individual application programs.
Gather data for each main transaction, and average values for the
total system. This includes the following data:
- Program calls per transaction
- CICS storage GETMAIN and FREEMAIN requests (number and amount)
- Application program and transaction usage
- File control (data set, type of request)
- Terminal control (terminal, number of inputs and outputs)
- Transaction routing (source, target)
- Function shipping (source, target)
- Other CICS requests