JVM in DB methodologies

Installing IBM's DB2b6.1 I notice that what appears to be a "standard" JDK and JRE are installed in the SQLLIB directory. A JAVA -fullversion reports that this JDK is the IBM 1.1.7 build, and it appears to be the same code that is from the direct JDK installation on my workstation.

Oracle claims to have their version of the JVM "embedded" in the database (JServer), meaning it is part of the server code, simply installed along side. But this is also a purchased option.

I can see benefits to each method. In the IBM method, the JVM appears to be "user-serviceable". I could (in theory) simply install IBM's 1.1.8 build when it is released. Given the fact that IBM is building it's own versions of the JDK, and the excellent performance with these versions, I see this as an advantage. Plus, I can interact with this VM as any other, no proprietary methods here.

In Oracle's product information pages on 8i, they indicate that their method of embedding the VM in the database code, allows their product to serve as a database and application server in one. They also state that their VM was engineered to be server specific, and thus has a performance advantage. However, their current JDK version seems to be at 1.1.6, and I have no way to directly compare the performance of the Oracle VM against IBM's.

Can anyone comment as to the pros/cons to either methodology? Received on Thu Jun 24 1999 - 15:43:38 CDT