0 Replies Latest reply on Nov 27, 2006 5:09 AM by gshriki

    Clustered-cache memory utilization issues

    gshriki

      Hi there,

      I?m experiencing some wired behavior in the JBoss Cache, regarding its memory management.

      Please have a look on the following scenario:
      1. Machine A creates a cache (a-sync mode).
      2. Machine A creates 50,000 objects (Each ? 1K).
      3. Machine B joins the cache.
      4. Machine A puts all 50,000 objects to the cache.
      5. Machine B gets all nodes to its local memory.
      6. Kill the application on machine B ().
      7. Machine B start the application again, and joins the cluster.
      8. Kill the application on machine B ().
      9. Machine B start the application again, and joins the cluster.
      10. Kill the application on machine B ().
      11. Machine B start the application again, and joins the cluster.
      12. Kill the application on machine B ().
      13. Machine B start the application again, and joins the cluster.


      For each step, I?ve measured the time (in MS) and the memory utilization.

      Some of the results:
      Average time to insert the nodes (by machine A) - 19,842.60 ms
      Average time to recieve all the nodes (by machine B) - 30,673.60 ms
      Average time to machine B to start (with loading all the cache data) - 18,954.50 ms
      Average memory usage to empty cache - 87.11 mb
      Average memory usage by objects (not in the cache) - 198.40 mb
      Average memory on machine A after step 5 - 550.40 mb
      Average memory on machine B after getting all data - 400.40 mb
      Average memory on machine A after step 7- 859.40 mb
      Average memory on machine A after step 8- 908.20 mb
      Average memory on machine A after step 11- 926.20 mb
      Average memory on machine A after step 13- 931.00 mb

      So my issues:
      It appears that each time machine B restarts, the memory utilization on machine A gets higher?
      Mainly I?m concern about the first time machine B stops and restarts. It takes the memory utilization on machine A from 550.40mb to 859.4mb?
      Another issue is the time that machine B takes to load all cache? ~19 sec. Is that what I need to expect?

      BTW, I?m using twi machine, both with Linux Red-Had Advance server 3 rc5. I'm using the top command to monitor the memory utilization for the processes.

      Thanks,
      Gilad.