2 CPUs of the following configuration(Intel Dual Core XEON Processor 3.3 Ghz )
Using thread dump on tomcat java process is a bit risky for us as it will bring down all our web services. On Mon, Sep 20, 2010 at 6:27 PM, Ashish <[email protected]> wrote: > How many CPU's/Cores on the machines? > > Try taking a few Thread dumps and see what's taking up the CPU. > > Alternatively, if its not a production App, go for profiling and see > what's going on inside. > > thanks > ashish > > On Mon, Sep 20, 2010 at 6:07 PM, Kumar Abhishek <[email protected]> wrote: > > Hi, > > > > I am using MINA for client-server based communication on my Centos > machine > > both of which reside on same machine. The architecture is the following: > > > > Client has a single SocketConnector and manages a pool of active > IOSession > > objects which can be utilized by any incoming request. This > SocketConnector > > is invoked by tomcat once HTTP POST request comes to a particular url > hosted > > on the machine. The client picks up an active session and fires the > request > > to the socket server. I am using executor Executors.newCachedThreadPool() > so > > that the request doesn't execute on IO thread. > > > > On the server end I have a socket server which uses a executor to process > > incoming IO request and send back response. The typical request > processing > > involves db queries, sending sms on smpp server and returning back the > > result to the client. Typically this request processing can take upto > 40secs > > to finish up(yeah I know its slow as sms sending and delivery report is > to > > be proceesed) > > > > What I observe is the following even on a single request the CPU usage of > > tomcat java process shoots up very high which is very worrying given the > > system is expected to see more traffic in coming days. I have also tried > to > > forcefully bring down the request processing time to 3secs by ignoring a > few > > steps just to diagnose the problem, however there is no improvement, its > > just that cpu shoots for a smaller period i.e. from 40 secs to 3 secs. > Any > > suggestions will be highly appreciated. > > > > Thanks ! > > Best Regards, > > Abhishek > > > > > > -- > thanks > ashish > > Blog: http://www.ashishpaliwal.com/blog > My Photo Galleries: http://www.pbase.com/ashishpaliwal >
