You may find the following performance measurements helpful in determining
the performance of a system:
- 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 (these can include
an intelligent cluster controller such as the 3601 and also the 3270 cluster
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 of time. 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 on the current
load on the network or DASD subsystem.
- Indications of internal virtual storage limits:
These 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. Paging rates to expanded
storage are only as important as its effect on processor usage.
- 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.
You should investigate both system conditions and application conditions.
A knowledge of these conditions enables 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.
These conditions, measured both for individual transaction types and for
the total system, give you an estimate of the behavior of individual application
programs.
You should gather data for each main transaction and average values for
the total system. This data includes:
- Program calls per transaction
- CICS storage GETMAINs and FREEMAINs (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.
[[ Contents Previous Page | Next Page Index ]]