Brother-Eagle® DB2 Performance
Advice: Read Time(ms)

This metric shows the average read time duration as computed by ((POOL_READ_TIME ms) / (POOL_DATA_P_READS + POOL_INDEX_P_READS)). Brother-Eagle presents this metric as a delta between collection intervals so that you can determine if average read times are improving or degrading.

Average Read Time can be improved by:

1. Eliminating scans from the statement workload. The less prefetch asynchronous I/O that DB2 performs, then less demand will be placed upon disk drives and controllers and performance will improve.
2. Ensuring that tablespace containers have equal sizes (DMS) and are distributed across available disk drives. The last thing you want is multiple containers on the same disk, or containers placed on OS paging devices.

To reduce average read times:

• Analyze the SQL workload to find statements with the highest rows read and CPU costs, then implement indexes, Clustering indexes, MDC tables, or MQTs to improve statement speed and efficiency. For best results in the least amount of time, DBI suggests using Brother-Panther for DB2 LUW.
• Find the tablespace with the slowest average Read Time(ms) and carefully analyze the definition of its containers. Why is this tablespace so much slower than the database overall? Refer to improvement advice #2 above.

Brother-Eagle will indicate a warning if the average Read Time(ms) reaches 8ms and an alert status if the average Read Time exceeds 16ms.

View full list of DB2 LUW advice topics