3 Replies Latest reply on Oct 11, 2010 8:24 AM by michal_szymanski

    Peroformance in cluster - where is bottleneck?

    michal_szymanski

      Recently I've tried to evaluate Infinispan performace using Cache Benchmark Performance (http://sourceforge.net/apps/trac/cachebenchfwk/wiki/FiveMinutesTutorial) but I've failed, during benchmark run I got strange error. It was the reason to write very simple test that puts 100 thousends request in one node and objects are replicated to other node (sources are here http://community.jboss.org/thread/157354?tstart=0).  Our nodes are based on  Linux with Core2Duo/4GB of RAM and 2.5"harddrive and we have used synchonous replication. During test I've got about 1500 puts/sec what is enough for our project but I wanted to know why performance is not better than 1500 and to be honest I still do not know. During test load of CPU/hard drive/network is very small, CPU load was aroud 20% per node, network traffic about 2MB/s (we used 1Gb network) and there was almost no hard drive operation. The question is where is performace bottleneck?  Any suggeston?