I want to quantitatively evaluate the performance of web service, i.e. measure the time difference from service invocation to the response. In the first case, I found that each client request response using java took abt 1sec which is very slow. For this test, the soap server is located on another machine in the same subnet and the client is my local PC. This time is the same regardless of whether I point to a DNS name or a resolve IPv4 no. For the second case, when I change my endpoint to localhost on my client and using a TCPMON(the stuff that lets you see the SOAP messages) to direct my calls to the foreign server, the invocation took very much less time. (about 0.3s). For the third case, I performed a localhost client invocation on the server machine itself (no TCPMON), it is very fast (<0.1s). Can anyone smart enough explain the why I am getting this large difference in invocation times? Is it a Apache stub implementation problem? Note: I would think this is definitely not a network latency problem. It does not make sense that the lag is due to resolving DNS in the JVM during each SOAP call as the timings were the same as in the first case.