https://bz.apache.org/bugzilla/show_bug.cgi?id=63916
--- Comment #27 from Mark Thomas <ma...@apache.org> --- I have been mixing up socket.txBufSize and socket.appWriteBufSize. My testing shows that socket.appWriteBufSize has no measurable impact on performance of the stand-alone test. Performance of the stand-alone test is tightly linked to socket.txBufSize. The smaller the buffer, the more likely it is the writing thread will fill it before the OS thread writes some data. That will cause the socket to be placed in the Poller and polling does appear to be relatively slow. The log messages are a significant factor as writing them to stdout slows things down. I've commented them out for now in my stand-alone test. The Tomcat unit test in question deliberately sets the txBufSize small to trigger the use of the Poller. It appears that that setting is far too small. I'll do some tests and increase it - probably to around 1MB. -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
