How Db2 gathers checkpoint information

At the group buffer pool checkpoint, the structure owner records in the SCA and in its own BSDS, the LRSN from which group buffer pool recovery should take place, if necessary.

This LRSN is displayed in the DISPLAY GROUPBUFFERPOOL command output. Db2 has two possible ways of gathering checkpoint information:

By issuing many read directory info requests
This method is used when the Parallel Sysplex® is not at the maintenance level required for the more efficient method described below. These read directory info requests are reported as an increase in the SRB time of the database services address space (ssnmDBM1), especially for the structure owner. This is because of the increased number of times that Db2 has to read the directory entries to compute the recovery LRSN.
By issuing one read castout statistics request
This method is used when the group buffer pool is allocated in a coupling facility at CFLEVEL=5 or higher.

When you look at the MDETAIL report from a DISPLAY GROUPBUFFERPOOL command you will see significantly fewer read directory info requests when you have the proper maintenance applied for this feature.

Recommendation: Because group buffer pool checkpoint consumes processor, coupling facility, and I/O resources and can impact other work in the system, balance the performance impact of frequent group buffer pool checkpoints (the lower the checkpoint interval, the higher the system resource consumption) with the recovery impact of infrequent checkpoints (the lower the checkpoint interval, the faster Db2 can recover from a group buffer pool failure). The default checkpoint interval of 4 minutes is a good balance between the performance and recovery considerations in most cases.