TCP delay on same host

Question

Hi there. I have two TCP applications running on the same host and one app needs to periodically send messages to the other at very short intervals. I am noticing an 80 to 100 millisecond delay from when the sender does a send() and the receiver's select() indicates the message has been received. It is very important for our application that there be as little latency as possible so 80-100ms is way too much. I cannot understand why this is so because as I said both processes are on the same host.
Do I need to tune some TCP parameters or do something special in setting up the sockets to avoid this delay? Any help or hints would be greatly appreciated!
[In case it helps: 1) if a message is transmitted after a gap of 1 second or greater, the receiver gets the message immediately without the latency mentioned above  2) the two processes are binding to the IP address of their host and not "localhost" or INADDR_ANY].
Thanks a lot in advance.
Sam.

Answer 1

hi gp and thank you very much for responding to this unusual problem.
- switch ports to the PCs are configured as portfast.
- switch ports between two catalyst switches are not configured (default)
- i didnt use the 'switchport access' command since they are default layer 2 interfaces. would 'switchport access vlan 1' command make any difference?
- i looked at the port status and confirmed connection is 100 mbps full duplex.
unusual issue is; ping, udp, multicast shows up in a very short time after I re-plug the uplink. that proves all ports are in forwarding state. only TCP shows up with delay, which doesnt occur on 200 $ unmanaged switch??
thanks in advance for any suggestions