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.

4.1.1 Per<strong>for</strong>mance<br />

support caused almost as much CPU degradation as the 64-bit conversion, and support <strong>for</strong><br />

variable length indexes was not far behind.<br />

Hopefully you can now appreciate how the new 64-bit architecture, combined with all the<br />

enhancements in <strong>DB2</strong> V8, have added much more code than V7 and have also presented<br />

some significant per<strong>for</strong>mance challenges which have had to be overcome in order to minimize<br />

the CPU regression of V8 compared with V7.<br />

We reviewed the per<strong>for</strong>mance of <strong>DB2</strong> V8, compared with <strong>DB2</strong> V7, <strong>for</strong> a variety of different<br />

workloads, to understand the impact V8 may have on CPU <strong>for</strong> your workload. The<br />

per<strong>for</strong>mance measurements we compared <strong>for</strong> each workload were the ITR (Internal<br />

Throughput Rate, inversely proportional to CPU seconds used) and CPU.<br />

We begin our discussion by having a look at OLTP workloads. The variety of OLTP workloads<br />

we studied include the IRWW (<strong>IBM</strong> Relational Warehouse Workload, described in <strong>DB2</strong> <strong>for</strong><br />

MVS/ESA <strong>Version</strong> 4 Data Sharing Per<strong>for</strong>mance <strong>Topics</strong>, SG24-4611), both in a data sharing<br />

and non-data sharing environment, a number of DRDA client/server transaction workloads, a<br />

number of SAP workloads and a CICS/<strong>DB2</strong> transaction workload. Each workload studied had<br />

a different SQL structure and there<strong>for</strong>e revealed different per<strong>for</strong>mance strengths and<br />

weaknesses. For example, some workloads executed reasonably simple SQL while other<br />

workloads executed more complex SQL.<br />

A number of query-based workloads have been studied to compare query per<strong>for</strong>mance. Over<br />

200 queries gathered from a number of customer sites have been used. Some of these<br />

queries achieved up to 20% reduction in CPU, due primarily to improved access path<br />

enhancements. For example, a data warehouse application achieved over 30% improvement<br />

in CPU, primarily due to star join enhancements.<br />

A number of batch workloads were also studied, including a sampling of commonly used<br />

utilities.<br />

Figure 4-1 summarizes some general measurements of V8 CPU consumption compared with<br />

V7 <strong>for</strong> various workloads. These comparisons are made with no application changes to<br />

exploit any V8 new function, nor any aggressive configuration changes.<br />

This chart has been compiled from a number of different laboratory tests running various <strong>DB2</strong><br />

workloads. A ‘+’ means a CPU increase and a ‘-’ means a CPU decrease in V8 compared<br />

with V7. Be reminded that these trends can only be used as a guide. As always, “your<br />

mileage may vary”, since all <strong>DB2</strong> installations and workloads are different.<br />

130 <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>

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

Saved successfully!

Ooh no, something went wrong!