IBM Preps New DB2 Feature to Improve Performance for Power Systems

 
 
By Brian Prince  |  Posted 2009-10-09
 
 
 

IBM is developing a new, optional software technology for IBMPower Systems to help organizations increase database transaction capacity while cutting the cost and risk associated with growing systems.

 

The new feature is called DB2 pureScale, and is slated to be ready for Power 550 Express and Power 595 in December. Developed jointly by the IBM Toronto Software Lab and Power Systems Lab in Austin, the technology is designed to help companies scale out their DB2 clusters, and as such, the feature competes against Oracle's Real Application Clusters technology. The software incorporates the new PowerHA pureScale technology to reduce the amount of communications required within the system, thereby slashing the amount of computing power wasted on overhead.

 

 "Each server in the cluster can interact directly with the PowerHA pureScale software through a very efficient InfiniBand network using Remote Direct Memory Access (RDMA)," explained Bernie Spang, IBM's Director of Product Strategy for Information Management. "This approach results in far less intra-system communications and eliminates costly IP network overhead. The payoff is a system where the highest level of computing power is applied to running database transactions and not wasted...It also delivers continuous availability and more consistent application performance than alternative designs, should there be a node failure."

  

 According to IBM, capacity growth achieved with DB2 pureScale does not require changes to applications or database tuning. The company added that this application transparency reduces risk and cost and also complements the capability delivered in DB2 9.7 that enables applications originally written for other database software, such as Oracle Database, to leverage DB2 with little or no change.

  

 "The key to real cost savings in a scale-out database environment is the delivery of true transparent application scaling," Spang said. "Scale-out architectures that rely on locality for effective scaling require developers make their applications cluster aware. Cluster-aware applications are more complex and costly to develop and deploy, and also require application rework when the cluster changes. Transparent application scaling means that applications do not have to be cluster aware in order to take advantage of the scale-out architecture."

 

"In essence, with pureScale, customers can simply scale the capacity of their existing systems without the need to purchase excess hardware and software - nor have to buy an overpriced, one-size-fits-all configuration that doesn't scale well and doesn't fit their needs," he said.

 

 

 

 

Rocket Fuel