1 Reply Latest reply: Jun 19, 2012 10:34 PM by Sridhar Bharadwaj RSS

    TCP Bottleneck(?) Resulting in HL7 Message Loss

      I am running SOA Suite across a 5-node cluster for the purposes of receiving HL7 messages via B2B. I am finding that under load many messages are failing to get processed by B2B, and the most likely culprit (that I can tell) is the managed server/cluster configuration. I am guessing that the number of inbound connections is far exceeding some parameter, resulting in the number of messages processed being far lower than what my sender says it sent.

      In doing some digging, it appears the message packets are making it to the server based on O/S stats. I also ran some independent tests on another listener platform and everything got picked up, so this seems to support my theory.

      Without going into a lot of specifics, our inbound traffic rates during the day are very high. CPU utilization is quite low, only around 25%. We have no leakage overnight or on weekends, when traffic rates are very low anyway.

      Where should I look to start opening things up a bit? Any help is appreciated.