“Staggering” Transaction Rates With DB2 pureScale

DB2 pureScale was first announced back in October 2009 and created huge interest in the market. In April of this year Triton sent a team of their experts to the IBM labs in Germany and were the first European IBM Business Partners to work with DB2 pureScale. Since then, the team have been working hard to build their skills with DB2 pureScale.

 

Although DB2 pureScale is only currently supported on IBM X Servers and Power Systems, Triton have shown that it is possible to implement DB2 pureScale on a low cost commodity cluster.

Listen to the podcast

 The purpose of creating the commodity cluster was to undertake some basic validation of IBM’s performance and scalability claims and assist IBM with early beta testing. Triton also wanted to build internal technical experience in a DB2 pureScale environment and establish a platform for ongoing research and development.

 

“Due to budget constraints and a need to make our cluster portable for customer demonstrations we had to keep the cost down to under £1000. The fact that we were able to do this and keep performance levels so high is encouraging. We have been truly astounded at the levels of performance we have seen with DB2 pureScale. It is a major step forward for DB2 on distributed systems.” James Gill, DB2 Expert, Triton Consulting.

 

The cluster has 2 member nodes and one Coupling Facility. Each node consists of:
• Intel D510M0 (Dual core 1GHz Atom)
• 4GB RAM
• 40GB SDD
There is a shared disk with DB2 9.8 pureScale FP2 development image. We used the Technology Explorer for workload and monitoring.

 

With this architecture we were able to drive 1000 transactions per second in a 2.5 million row table. When considering that there was absolutely no tuning done on this, it was straight out of the box, these are extraordinary figures.

 

We then went on to implement the IBM NanoCluster which is a 2-host cluster configuration with a CF (Coupling Facility) and member on each node. The NanoCluster comes with a well-tuned database. Using this configuration we saw a staggering 6000 transactions per second!

 

You can see a range of live demonstrations which we recorded with the IBM NanoCluster on You Tube:
DB2 pureScale dynamic licensing and capacity

DB2 pureScale availability demonstration – CF Failure

DB2 pureScale automatic member recovery

For more information on Triton’s DB2 pureScale implementation and the IBM NanoCluster listen to the podcast with James Gill by clicking on the link at the top of this article.