DB2 – faster than a speeding bullet?

DB2 – faster than a speeding bullet?

Last week, IBM sent out press releases saying that it has achieved the industry's highest ever TPC-C (transaction processing) benchmark using a Power Systems configuration with DB2, hitting 10,366,254 transactions per minute. And in case you don't realise just how fast that is, the press releases goes on to inform us that it beats HP's best result by more than 2.5 times and Oracle's best by more than 35%!

We're told that: "The results place IBM in a unique position as the undeniable leader. With these new clustered results (and with the long-standing single system result), IBM has demonstrated its ability to scale up to handle higher transaction loads and to scale out to optimize more types of workloads than the competition."

This result is the largest ever TPC-C result published (dated 17 August 2010) and was obtained on a cluster of three IBM Power 780 servers featuring a storage subsystem with 116TB of Solid State Drives (SSDs) and running DB2 9.7.

Just to rub in the fact that it beats the Old rival, Oracle, the press releases goes on to inform us that: "The IBM result represents 2.7 times better performance per core than the Oracle result, 41% better price performance, and 35% better energy efficiency per transaction. IBM's performance is also more than 2.5 times better than HP's best result, 69% greater performance per core, and 2.1 times better price/performance."

The new TPC-C benchmark result uses standard IBM software. DB2 9.7 has been around since June 2009, and AIX 6.1 was released in November 2007. IBM says that the selection of software versions reflected levels currently in use by a large number of their clients.

IBM highlights the fact that the IBM TPC-C results on POWER7 technology shows off IBM Storage technology in the form of Solid State Drives (SSDs), which enable higher throughput and lower response times. SSDs also provide reliability, lower energy usage, less cooling requirements, and the ability to reduce data centre footprints. The total storage used was over 800TB while the Oracle/Sun configuration had 686.6TB of total storage.

The Oracle/Sun cluster is 71% more expensive based on published price/performance relative to the IBM Power 780 cluster result. The total system cost of the Sun cluster is 26% greater than the total system cost of the IBM cluster. Because of different discount structures, care should be taken in comparing individual price components.

As a result of these differences, the TPC does not allow comparisons using TPC price information on anything other than the total configuration. The IBM Power 780 with DB2 9.7 result yields greater than 10 million tpmC for IBM (a feat Oracle hasn't accomplished) and significantly better price/performance for the IBM solution.

The configuration for this benchmark achieves an estimated consumption of 65.1 kWatts or 6.3 kWatts per million tpmC, 35% better than the Sun cluster energy consumption estimate of 73.9kWatts or 9.7 kWatts per million tpmC.

You get the idea anyway – there's more of the same in the press release. The bottom line is that IBM can now claim to provide performance in excess of double-digits (10 Million Transactions Per Minute), which, they say, no-one else can do.