Capping in a single logical partition

In order to use capping for an LP on a CPC where there is a need for only one active LP using shared CPs, you must define and activate a second dummy LP. The dummy LP must also be defined as using shared CPs. The weights of the two LPs can be adjusted to attain the desired cap for the one LP that will actually be used.

The dummy LP does not have to be capped. In most cases, the dummy LP does not need to have anything IPLed into it. If nothing is IPLed into the dummy LP, the dummy LP will not use any CP resources. In some situations where the single capped LP runs an extremely CP-intensive workload, it may be necessary to run a program in the dummy LP to smooth distribution of CP resources to the LPs. The program can be a simple branch loop that spins to waste time. Without this program running in the dummy LP, the CP-intensive, capped LP can experience a lurching effect with its dispatched times. It will be capped to its weight properly, but it could get all of its allotted CP resource quickly and then wait for a period of time before it can run again.

If nothing is to be IPLed into the dummy LP, the dummy LP can be defined and activated with no channel paths or devices. A definition with no devices is desirable to prevent control unit logical paths from being established for the dummy LP. See Managing logical paths for FICON channels.

As alternative to this procedure for capping a single logical partition, consider using the absolute capping support described in Absolute capping or capping a group of logical partitions described in Absolute group capping.