07.02.2013 Views

Best Practices for SAP BI using DB2 9 for z/OS - IBM Redbooks

Best Practices for SAP BI using DB2 9 for z/OS - IBM Redbooks

Best Practices for SAP BI using DB2 9 for z/OS - 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.

11.4 Test case scenarios<br />

The goal of the Jupiter project was to prove the scalability of a <strong>BI</strong> Accelerator <strong>for</strong><br />

increasing data volumes and increasing blade resources. The test plan consisted<br />

of a suite of tests on three database sizes (5 TB, 15 TB, and 25 TB), various<br />

blade configurations (27, 81, and 135 blades), the <strong>BI</strong> Accelerator index creation,<br />

and a multi-user query workload.<br />

Note that <strong>for</strong> every 27 blades, one backup blade was configured. This brought<br />

our total blade requirements to 28, 84, and 140 <strong>for</strong> the three database sizes.<br />

11.4.1 <strong>BI</strong> Accelerator index creation<br />

When all the data was populated into the InfoCubes in a <strong>DB2</strong> database on<br />

System z, the next step was the <strong>BI</strong> Accelerator index creation. This phase read<br />

data from a row-based database and trans<strong>for</strong>med it into in-memory<br />

column-based files in a cluster of blades. In the Jupiter study, all InfoCube data<br />

from a <strong>DB2</strong> database was indexed to <strong>BI</strong> Accelerator blades be<strong>for</strong>e it was<br />

available <strong>for</strong> query processing.<br />

Objectives<br />

The purpose of the <strong>BI</strong> Accelerator index creation test scenarios was to measure<br />

key per<strong>for</strong>mance indicators (KPIs) on 5 TB, 15 TB, and 25 TB database size<br />

points with blade resources of 27, 81, and 135 systems, respectively. These<br />

KPIs were:<br />

► Elapsed time to create indexes into the <strong>BI</strong> Accelerator<br />

► <strong>BI</strong> Accelerator index creation throughput rate<br />

► CPU utilization on <strong>BI</strong> server and blade servers<br />

Description<br />

Since the goal of <strong>BI</strong> Accelerator index creation is to read data from a database<br />

server to a GPFS on blades, certain sizing and configuration settings need to be<br />

considered. Proper <strong>BI</strong> Accelerator blade memory sizing is an important<br />

consideration. <strong>SAP</strong> note 917803 provides an official sizing estimate and<br />

guideline <strong>for</strong> blades. In this study, 135 blades plus five backups (10 HS21<br />

chassis of 14 Woodcrest blades each) were used. Each blade had 16 GB RAM<br />

and two dual-core processors at 3 GHz. Thus, 140 blades had about 2.2 TB of<br />

memory (RAM). A rule of thumb is to allocate 50% of blade memory <strong>for</strong> storing<br />

the indexed cube data and 50% <strong>for</strong> the Linux system <strong>for</strong> processing runtime<br />

objects and storing query result sets. This avoids per<strong>for</strong>mance impacts from<br />

memory swapping to disk.<br />

238 <strong>Best</strong> <strong>Practices</strong> <strong>for</strong> <strong>SAP</strong> <strong>BI</strong> <strong>using</strong> <strong>DB2</strong> 9 <strong>for</strong> z/<strong>OS</strong>

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

Saved successfully!

Ooh no, something went wrong!