Problems passing large messages >2MB from site to site

Steve,

I can’t think of a reason why Cloverleaf would cause the message to go slow.  It could be that the stack is breaking the message down into many transmissions though.  It might be interesting to see what the packet size is for tcpip on your system.  It may be possible to see this in the log if you use enable_all on the destination process.  Take a look at the packet hex dumps in the log to see how long they are.

Also, make sure that you push some small messages and see that they go fast or slow.  That will at least tell you if it’s systemic or based on large messages.

We have pushed small messages through and were able to process to the tune of between 20-30 messages per second.  We have also set up a socket connection between the two sites using nothing but TCL in the outbound TPS (not a great option for production but we wanted to see what it would do) and found that the socket communications itself is not a bottleneck.  I assume (hate to say that) that there is some sort of configurable parameter that we can modify that will give us access to the socket options in order to tune things.  I say this because the TCL allows use to use fconfigure to set parameters and an external java application would allow us to call a setSockOptions on the socket.

If there is not a configurable parameter of some sort for this purpose, we are going to end up performing this function outside of cloverleaf and I would REALLY like to avoid that.