Hi Ralph > We repeatedly had problems in the case that the partner's backend dropped > out or got slow: > The portlet opens more and more connections to its backend that get stale > immediately. > Do you know if HTTP keepalive connections were used? also what the TCP socket timeout at the two ends - i.e. the portal and the backend? Either way to overcome this issue, we need to establish how to keep talking to the partner backend in the most efficient way - and the NIO transport is capable of doing that quite very well. So I am confident of a solution being within reach.. but need to consider a bit more. > This problem extends until the portal server system cannot open any more > connections and drops down too. > Have you tuned the OS to increase the available sockets on the client? Usually when we run a load / perf test with the ESB, there is a minimum OS level configuration we perform for best performance. > In order to overcome this, we are thinking of placing some kind of shield > between the portlet and the backend. > One idea is to place an esb in the middle and limit the size of > connections to and from the esb. > Although this is not controlled by Synapse, it is possible to do this with a bit of coding at most - however, I believe that we may not need this if we properly tune the ESB to partner backend. BTW, is this a public endpoint
-- Asankha C. Perera AdroitLogic, http://adroitlogic.org http://esbmagic.blogspot.com
