Examining paging rates

The operating system allocates real storage for the address spaces occupied by the control region or dependent regions. IMS might not obtain adequate machine cycles because of the excessive demand for real storage. Observe the paging rates and the frequency with which processing is delayed by a page fault. You can use the output from RMF II to examine the paging rates at peak IMS loads.

Page faulting in an IMS online system can directly increase the transaction time-in-system by the duration of the paging activity. If page faulting must occur, it is best to confine it to application programs in the dependent regions. A page fault experienced by the control region can hold up any dependent region waiting for a control region service. IMS provides some page-fixing options; these are covered in Page fixing the IMS buffer pools. However, if real storage is constrained, any page fixing generally serves to increase paging elsewhere and is not recommended. The primary tuning technique must be to reduce the virtual and the real storage requirement. This is particularly true in a dedicated IMS environment, where page fixing some portion of IMS only makes paging worse in the unfixed portion, possibly making overall paging rates worse.

You can monitor page rates dynamically, daily, or in detail. For more information about monitoring page rates, see Deciding on monitoring activities and techniques.

After the reason for the increased storage over commitment is identified, one or all of the following actions must be taken:

• Trade off IMS paging I/O against IMS controlled I/O.
• Reduce IMS paging I/O by reducing over commitment (reduce concurrent users) or by increasing IMS priority relative to other users.
• Speed up paging I/O by reducing I/O contention, splitting page data sets, or by placing page data sets on faster devices.
• In a non-dedicated IMS environment, try to reduce IMS paging, at the expense of other competing work, by fixing some portions of IMS.