Thanks a lot. It seems that's fully answers my question. We will use MTOM attachments in such case.
2009/9/30 Daniel Kulp <[email protected]> > > It does sound like you are becoming bandwidth or latency bound on the > remote > connections. My gut feeling is that on the local network, you are > becoming > CPU bound. The time it takes to escape the string is greater than the > bandwidth restrictions so MTOM is really winning out. > > On the remote connections, the bandwidth is effectively throttling things. > For mtom, the cpu ends up just sitting around doing nothing while it waits > to > send stuff. With non-mtom, the cpu can escape strings while data is being > sent. > > I'm willing to be that if you stuck a CPU monitor on, the MTOM case would > show > much lower cpu usage. That COULD be good from a scalability standpoint as > that CPU could be used to process additional requests and such. > > Dan >
