intersite communnication delays (backlog on outbound thread)

Clovertech Forums Read Only Archives Cloverleaf Cloverleaf intersite communnication delays (backlog on outbound thread)

  • Creator
    Topic
  • #53917
    Carson Nguyen
    Participant

      Fellow Clovertechers,

      I’d like to pick your brains to see what potentially could cause a performance issues/delays for intersite communications.  

      Here’s a somewhat detailed description:

      I have an outbound thread originating on a collection site (high volume) to disperse to other sites.  This thread along with a few others build up to a high OB queue depth at times.  But if you trace the transaction flow to the destination site, you won’t see an inbound queue depth.  Any ideas on potential bottlenecks?

      OB thread configuration (originating site):

      –no processing overhead within the process i.e. ob tps proc, xlate, etc

      tcp/ip with “mlp_tcp.pdl” set as client with a timeout specification of 60s for awaiting replies

      tps inbound reply is set to utlize “cl_check”ack”


      IB thread configuration (destination site):

      –TPS Inbound Data “STACK” consisting of: raw_hl7_ack, transaction filters, conversion scheme, another adt logic script (checked to run in a “sub-thread”)

      tcp/ip with “mlp_tcp.pdl” set as “Server”

    Viewing 1 reply thread
    • Author
      Replies
      • #79521
        Jim Kosloskey
        Participant

          Carson,

          Have you checked the delay between the sending of the message (outbound SMAT) and the receipt of acknowledgment (inbound SMAT) on outbound thread?

          Are you seeing any resends (do you resend messages on timeout)?

          In the receiving site are any of the routes or outbound threads constrained (deep queues)?

          email: jim.kosloskey@jim-kosloskey.com 29+ years Cloverleaf, 59 years IT - old fart.

        • #79522
          Elisha Gould
          Participant

            This will happen when there are delays in processing in the destination site. If there is only one process and one of the threads lag, there will be queues on your collection site, as the processing is all single threaded.

            We have had this issue when using the http-client for outbound on the destination site, as it can take several hundred milliseconds to call and return from Curl.

            To overcome this, on the destination site, we put the inbound thread in a separate process to the outbound thread.

            There are still queues, but only in the destination site.

        Viewing 1 reply thread
        • The forum ‘Cloverleaf’ is closed to new topics and replies.