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

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

CPU reduction using in memory work file<br />

CPU Time (s)<br />

600<br />

500<br />

400<br />

300<br />

200<br />

100<br />

0<br />

4232 4509 4416 4135 4457<br />

4580 3016 4417 4485 4901<br />

QUERY NO<br />

Figure 3-22 In memory work file - CPU reduction<br />

V8 NO IMWF<br />

V8 64 MB<br />

Per<strong>for</strong>mance improvements observed on queries from a workload with 64 MB of IMWF<br />

compared with no IMWF:<br />

► Up to 40% elapsed time reduction<br />

► Up to 31% CPU time reduction. Top queries from the workload:<br />

– Query 4580: CPU time NO IMWF = 560 sec. and with IMWF = 446 (reduction of 21%)<br />

– Query 4485: CPU time NO IMWF = 559 sec. and with IMWF = 386 (reduction of 31%)<br />

– Query 4901: CPU time NO IMWF = 553 sec. and with IMWF = 411 (reduction of 25%)<br />

In practical environments 128 MB to 256 MB can be a good starting point, depending on the<br />

workload using the snowflake data model.<br />

Bigger improvements are anticipated in environments where work file data sets are shared<br />

concurrently with other threads, such as <strong>for</strong> sorting.<br />

Per<strong>for</strong>mance study on warehouse workload<br />

Total elapsed time and total CPU time were measured <strong>for</strong> 79 star join queries in a highly<br />

normalized data warehouse workload in V7 and V8.<br />

Figure 3-23 shows the improvement in per<strong>for</strong>mance with sparse indexes, access path<br />

changes and the use of in memory work file.<br />

Chapter 3. SQL per<strong>for</strong>mance 63

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

Saved successfully!

Ooh no, something went wrong!