Version 2

    Hi,

     

    I'm trying to build a hot standby cluster application with 2 servers (primary and secondary) both link by back to back cable, data synchronization and heartbeat in between will implemented by using jgroups. And i'm trying to extend HSQL by using ReplicatedHashMap so that i can survive in case of hardware failure. I'm doing performance testing with

     

    Setup

    - my desktop P4 3.2G 4Gmem

    - loopback interface (127.0.0.1)

    - default tcp.xml configuration

    - jdk 1.6

     

    Case

    Server tester will wait for Client tester and try to put a 10000 unique Serializable object which have three fields (int, String and BigDecimal).

     

    Result

    On average ReplicatedHashMap take 0.57ms to complete in put and around 1750 object per sec.

     

    And my questions are

     

    Questions

    1. For primary and secondary setup which configuration is the best (e.g. tcp.xml or tcp-nio.xml) ?

    2. Any way that i can tuning up by changing configuration parameter?

    3. I did a profiling and seem that more of the time in put operation spent in TP.send, is it expected?

     

    Appreciate if you could throw some light on this, thanks a lot!

     

    P.S. Enclosed please find those classes for my performance testing and Eclipse execution time profiling call tree.

     

    Best Regards,

    From Kan