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.

Table 4-4 System storage configuration <strong>DB2</strong> V7 vs. <strong>DB2</strong> V8<br />

<strong>DB2</strong> V7 uses a data space <strong>for</strong> the dynamic statement cache. In addition, there is a negligible<br />

use of the Sort pool, RID pool, and global dynamic statement cache in the EDM pool (less<br />

than 1 MB). The parameter KEEPDYNAMIC was set to NO so there is no local dynamic<br />

statement cache.<br />

Here are the different types of workloads we examined to understand the impact of static vs.<br />

dynamic SQL:<br />

► Dynamic SQL using CLI and OBDC on AIX, connecting to <strong>DB2</strong> Connect and using DRDA<br />

to communicate with <strong>DB2</strong>.<br />

► Dynamic SQL using the <strong>DB2</strong> Universal Driver Type 4 on AIX, connecting to <strong>DB2</strong> Connect<br />

and using DRDA to communicate with <strong>DB2</strong>.<br />

► Static SQL embedded in a local application running on AIX, connecting to <strong>DB2</strong> Connect<br />

and using DRDA to communicate with <strong>DB2</strong>.<br />

► Static SQL using the <strong>DB2</strong> Universal Driver Type 4 on AIX, using DRDA to directly connect<br />

to <strong>DB2</strong>.<br />

► A local IRWW OLTP workload directly connecting to <strong>DB2</strong> using IMS attach.<br />

► A local IRWW OLTP workload directly connecting to a two member data sharing group<br />

running <strong>DB2</strong> using IMS attach.<br />

During the per<strong>for</strong>mance tests, RMF storage statistics were gathered as well as <strong>DB2</strong> storage<br />

statistics using IFCID 225. In order to validate the storage statistics reported by <strong>DB2</strong> PE, and<br />

storage in<strong>for</strong>mation collected from the RMF monitor reports, system dumps were taken of all<br />

<strong>DB2</strong> address spaces. In addition, SDSF DA (Display Active) real storage usage was<br />

captured. As a result, we were able to compare the difference between <strong>DB2</strong> V7 and V8 DBM1<br />

and DIST virtual and real storage per DBAT and active connection. The results are presented<br />

here.<br />

Remember that your results may be very different as most customer workloads do not<br />

resemble this configuration; usually customers have a mix of different applications all running<br />

in a single <strong>DB2</strong> subsystem. Take this into consideration as you examine the potential impact<br />

of your migration to <strong>DB2</strong> V8.<br />

Virtual and real storage comparison <strong>for</strong> DBM1<br />

We first look at virtual storage used by the DBM1 address space.<br />

Figure 4-4 compares the total virtual storage usage <strong>for</strong> the DBM1 address space <strong>for</strong> all of<br />

these different workloads with <strong>DB2</strong> V7 vs. V8. However, to be fair, the virtual buffer pools were<br />

removed from the equation; you can see that GETMAINed storage is represented by GM - BP<br />

in the graph where the storage <strong>for</strong> these buffer pools were removed.<br />

A significant decrease in storage below the 2 GB bar is possible. With V8, several DBM1<br />

address space structures that consume large amounts of storage below the bar, have moved<br />

148 <strong>DB2</strong> <strong>UDB</strong> <strong>for</strong> z/<strong>OS</strong> <strong>Version</strong> 8 Per<strong>for</strong>mance <strong>Topics</strong><br />

<strong>DB2</strong> V7 <strong>DB2</strong> V8<br />

BP type Primary (MB) Above 2 GB bar (MB)<br />

4 KB BPs 438 438<br />

8 KB BPs 16<br />

16 KB BPs 16<br />

EDM pool 146 146

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

Saved successfully!

Ooh no, something went wrong!