Oracle Shows How To Juggle Data For Application Performance On The Grid

Oracle is taking a recent acquisition, Coherence, and enhancing its ability to manage running applications. Being able to manage the dark arts of application performance may prove as important in the long run as supplying the application itself.

Coherence 3.3 will become available Aug. 13 as an added piece of Oracle Fusion middleware. With a price tag of $20,000 per CPU in its Data Grid edition, it is software that enables what Gartner refers to as an "extreme" transaction processing environment.

Coherence is in use at insurance supplier Geico.com, Putnam Investments, Wachovia Corporate Investment Bank, Federal Express, and CheckFree, according to Cameron Purdy, the former CEO of Coherence's former owner, startup Tangosol. Oracle acquired Tangosol in June, and Purdy became an Oracle VP of development. Coherence takes the data that an application needs for transactions and spreads it across the random access memories of servers in a grid or cluster. Instead of going to a storage disk, the application is able to tap data at a much faster speed directly out of server memories.

Coherence also manages the data in such a way that if a server in the cluster should fail, it can re-establish a transaction with a backup copy of the data on a different server. Recovering from system failure in the past has been dependent on rebuilding the transaction record from a server log file. Coherence is able to recover transactions at the much faster speed by rebuilding from data in server memories.

It may be an exaggeration, but Purdy calls the transaction at the moment of system failure "in a pre-recovered state on another machine." It sounds like a hot-mirrored setup, where a duplicate of a running system gets fed exactly the same data. But Coherence is meant to govern, not two closely matched servers, but hundreds or thousands of servers at a time.

Coherence-supported applications can meet increased traffic loads by assigning more servers to the application's work, spreading data across them as well. Its main effect is to give high availability to an application and supply it with more resources upon demand tapping more servers in the grid. It is typically running on commodity hardware when it does so, Purdy added.

With Release 3.3, Oracle has added self-tuning capabilities to Coherence, enhanced its ability to tap the processing capabilities of multicore CPUs, and introduced Deterministic Request Execution features. By monitoring transactions, Coherence knows when one has become locked up on some node on the grid. It traces the point at which it was stopped and reactivates it at the same point elsewhere to sidestep the system failure or application lock-up.