IBM 5695-DF1, 5655-DB2 Storage Server Analysis, Conclusions, RMF Views, Summary of I/O operations

TOT4K	TOTAL
---------------------	--------
BPOOL HIT RATIO (%)	2	E
GETPAGES	6135875
BUFFER UPDATES	48
SYNCHRONOUS WRITE	0
SYNCHRONOUS READ	19559	F
SEQ. PREFETCH REQS	164649	G
LIST PREFETCH REQS	0	H
DYN. PREFETCH REQS	26065	I
PAGES READ ASYNCHR.	5943947	J
HPOOL WRITES	0
HPOOL WRITES-FAILED	0
PAGES READ ASYN-HPOOL	0
HPOOL READS	0
HPOOL READS-FAILED	0

Figure 66. DB2 PM Accounting Buffer Pool Summary

12.1.4 Conclusions

This example shows a very complex and heavy read-only query. Five-way parallelism has been used to reduce elapsed time.

The query is CPU bound with complex stage 2 type predicates, which access many pages but produce a small result set of 100 rows.

Sequential prefetch, dynamic prefetch, and sequential reads are executed. I/O response time is excellent.

The query could probably benefit from a higher degree of parallelism or from a faster central processor.

Summary of I/O operations
Sequential prefetch requests:	164649
Dynamic prefetch requests:	26065
Total prefetch requests:	190714
Total synchronous reads:	19559
Total I/O requests:	210273
Average I/O per second:	93

12.2 Storage Server Analysis

From the storage server analysis point of view, the first activity, in such a case study, is to eliminate unnecessary data from large RMF and IXFP reports. After the data has been discarded, you apply the methodology described in Chapter 11.2, “RMF Monitoring” on page 124, and Chapter 11.3, “IXFP Monitoring” on page 135 to access the relevant information.

12.2.1 RMF Views

RMF generates a report set for each LPAR. In each set, the cache subsystem activity report displays the same common view over all sharing LPARs at the LCU level. Therefore, the process of extracting data to analyze is mainly based on

Case Study 147