I have run some simple performance tests comparing the 10.3.1.0 beta build against 10.2.2.0 to check the effect of the various performance improvements that have gone into 10.3. The first test measure the throughput of reading a random record from a table while the second test runs "TPC-B like" load (each transaction consists of three updates, one insert, and one select). The from these tests are attached.

As the graphs show, 10.3 contains large improvements in the performance, particularly for read-only load. For reading random records 10.3 has more than 3 times the performance of 10.2 (on this test configuration) when there are concurrent clients. Great work!!

The tests were run on a dual AMD Opteron machine running Solaris 10 and JDK 6.

Some more performance data comparing 10.3 to 10.2.2 can be found here:

 http://home.online.no/~olmsan/derby/perf/

This page shows the results from a daily run of trunk and compares the performance using 10.2.2. as baseline.

Olav

<<inline: select-mm.png>>

<<inline: tpcb-mm.png>>

Reply via email to