Integration Watch: Fusing a coherent Oracle
March 15, 2009 —
(Page 1 of 3)
Related Search Term(s): BEA, IBM, Oracle
When Oracle bought BEA Systems last year, many pundits viewed the purchase as a good move if Oracle could put together a sound strategy for integrating the various pieces BEA's product portfolio into something useful. There was no doubt it made sense to fill out the DBMS and SOA story with a full-blown enterprise Java EE server and a wide assortment of well-regarded middleware from BEA. The strategic benefit of the move was to give Oracle a far more compelling story with which to battle IBM.
Prior to the acquisition, Oracle’s Java EE server was OC4J, a nimble server with a small footprint but lacking the creds of enterprise capacity. In BEA’s WebLogic offering, Oracle acquired a server that was better head-to-head than IBM’s WebSphere. Not only is WebLogic faster in terms of performance, it’s also consistently earlier in adopting key standards.
Another component in the stack that Oracle lacked, compared with IBM, was the JVM. JVMs are widely underappreciated as contributors to performance, and they carry an extraordinarily high bar to entry. Acquiring one of the three proven commercial JVMs was a coup in itself. However, the BEA JVM, called JRockit, is in a class by itself compared with its two competitors.
To wit, effectively every third-party server vendor runs its SPEC benchmarks using JRockit, and like WebLogic, the JRockit JVM is an adopter of new standards and architectures. For example, it was the first 64-bit JVM to be released for the Intel EM64T/AMD64 architecture.
But what really sets JRockit apart is incremental garbage collection (GC). In the other JVMs, GC is non-incremental: It occurs at unpredictable moments and can last for indeterminate times, during which the JVM can do little other work. As a result, those JVMs are highly unpredictable in terms of actual performance.
JRockit does constant, incremental GC so that its performance is much closer to a deterministic model. This predictable performance is a desirable attribute in absolute terms, but invaluable for IT departments under the gun to deliver against tight SLAs.