performance compare

Hi,
Our Team performed a comparison test of the Index Tree Calculation implementation. The first implementation was in Weblogic Server and the second was in Java Stored Procedures. The comparison was done in terms of how much time it would take to complete the calculation. Before presenting the results, below is a brief description of each implementation.

Weblogic Server Implementation: In this case, the Index Tree parameters are retrieved from the Database through network calls, the Index calculation is done in the Weblogic Server and the results are stored in the Database. The computation times were measured in the Weblogic Server.

Java Stored Procedure Implementation: In this case, the Java code for data retrieval, Index tree calculation and data update are stored in the Database. When Index calculation is initiated, the Database starts a Java Virtual Machine within the Database, runs the code to retrieve data, calculates the indexes and updates the database with the calculated data. The computation times were measured in the Database.

In summary, we compared between the cases of running the calculation process outside the Database (network calls involved) and within the Database. The code that does the Data retrieval, Index calculation and index updates is the same (with some minor modifications that did not affect the calculation process) in both implementations.

Environment

The Weblogic Server was running on a PC and the Database was running on server PPI2. The time for calculation was measured in milliseconds. The calls were made to the database in both implementations using JDBC.

The Java Virtual Machine (JVM) running on the Database Server (to run Java Stored Procedure) was fine tuned by the Database group.

Result Analysis:

The Weblogic Server implementation fared well compared to the Java Stored Procedure implementation in our testing. Weblogic Server implementation took 1.8 seconds for calculating multiple trees compared to 13.9 seconds for Java Stored Procedure.

Before we did the performance comparison, we expected the Java Stored Procedure to perform better than the Weblogic Server implementation. Our premise was based on the network calls being made from the Weblogic Server to the Database to build the tree details before calculation. But the Weblogic Server implementation provided a better performance than the Java Stored Procedure implementation. Could you advise us?

Thanks much!

CH Received on Tue Aug 26 2003 - 09:09:41 CDT