19.06.2013 Views

DB2 UDB for z/OS Version 8 Performance Topics - IBM Redbooks

DB2 UDB for z/OS Version 8 Performance Topics - IBM Redbooks

DB2 UDB for z/OS Version 8 Performance Topics - IBM Redbooks

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

► Configuration<br />

– <strong>DB2</strong> <strong>for</strong> z/<strong>OS</strong> V8 NFM<br />

– Non-data sharing<br />

– z/<strong>OS</strong> Release 1.5<br />

– DFSMSdss V1R5<br />

– <strong>IBM</strong> z900 Series 2064 4-way processor<br />

– ESS 800 DASD<br />

– FlashCopy V2<br />

► Workload <strong>for</strong> measurement<br />

– 20 M rows, 6 indexes, 1 table space with 10 partitions<br />

► Measurement scenarios<br />

– CHECK INDEX SHRLEVEL(REFERENCE), CHECK INDEX SHRLEVEL(CHANGE)<br />

against<br />

Partitioned Index<br />

Non-partitioned index<br />

All indexes (6) of the table space<br />

– Concurrent CHECK INDEX SHRLEVEL(REFERENCE) PART of all partitions of the<br />

table space versus single CHECK INDEX SHRLEVEL(CHANGE) against<br />

Partitioned index<br />

Non-partitioned index<br />

CHECK INDEX measurement results and study<br />

We look at several measurements using and not using FlashCopy, and executing CHECK<br />

INDEX on concurrent partitions.<br />

SHRLEVEL CHANGE with FlashCopy vs. SHRLEVEL REFERENCE<br />

This test case measures the CPU and elapsed time to compare CHECK INDEX with option<br />

SHRLEVEL CHANGE using FlashCopy V2 to CHECK INDEX with option SHRLEVEL<br />

REFERENCE. The results of this test case are shown in Figure 6-2. SHRLEVEL CHANGE<br />

using FlashCopy reduces the elapsed time dramatically.<br />

The observations of this study are as follows:<br />

► With the SHRLEVEL CHANGE option, <strong>DB2</strong> uses parallel tasks in the execution of the<br />

utility. In the case of a partitioned index (PI), the UNLOAD, SORT and CHECKIDX tasks<br />

are processed in parallel in each partition of the table space. In the case of a<br />

non-partitioned index (NPI), the UNLOAD and SORT tasks are processed in parallel, like<br />

<strong>for</strong> PI, but the CHECKIDX and MERGE phases do not use parallel processing.<br />

► Most CPU time improvement in SHRLEVEL CHANGE of partitioned index comes from the<br />

CHECKIDX phase. Also SHRLEVEL CHANGE of non-partitioned index improves CPU<br />

time in the CHECKIDX phase in comparison with SHRLEVEL REFERENCE, but the total<br />

CPU time increases in SHRLEVEL CHANGE because of SORT and MERGE phases. The<br />

MERGE phase exists only in the case of a non-partitioned index and SHRLEVEL<br />

CHANGE.<br />

If we only look at the CHECKIDX phase, we can say that SHRLEVEL CHANGE improved<br />

CPU time significantly both in PI and NPI. SHRLEVEL CHANGE makes the total elapsed<br />

time decrease dramatically both in PI and NPI. A PI can benefit from the per<strong>for</strong>mance<br />

improvement of online CHECK INDEX more than an NPI.<br />

► CHECK INDEX 6 indexes of the table space with SHRLEVEL CHANGE has 18 tasks.<br />

– UNLOAD 6 indexes in parallel<br />

Chapter 6. Utilities 265

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!