Solaris network configuration (poor performance)

Question

I am wondering if there is anything in the Solaris 10 configuration that I may be able to change to resolve a strange network performance issue I'm facing.

I have two Solaris 10 machines, both of which communicate with a 3rd. Between the first and third, I see the following network transfer performance:

me@sever_three:/var/adm> netio -t server_one

NETIO - Network Throughput Benchmark, Version 1.26
(C) 1997-2005 Kai Uwe Rommel

TCP connection established.
Packet size  1k bytes:  3652 KByte/s Tx,  3012 KByte/s Rx.
Packet size  2k bytes:  13977 KByte/s Tx,  10910 KByte/s Rx.
Packet size  4k bytes:  5604 KByte/s Tx,  14780 KByte/s Rx.
Packet size  8k bytes:  52212 KByte/s Tx,  30345 KByte/s Rx.
Packet size 16k bytes:  25432 KByte/s Tx,  38524 KByte/s Rx.
Packet size 32k bytes:  41940 KByte/s Tx,  52256 KByte/s Rx.

while to the second I see:

me@server_three:/var/adm> netio -t server_two

NETIO - Network Throughput Benchmark, Version 1.26
(C) 1997-2005 Kai Uwe Rommel

TCP connection established.
Packet size  1k bytes:  3804 KByte/s Tx,  400 KByte/s Rx.
Packet size  2k bytes:  16477 KByte/s Tx,  1349 KByte/s Rx.
Packet size  4k bytes:  76230 KByte/s Tx,  1384 KByte/s Rx.
Packet size  8k bytes:  33076 KByte/s Tx,  6782 KByte/s Rx.
Packet size 16k bytes:  108829 KByte/s Tx,  111466 KByte/s Rx.
Packet size 32k bytes:  112070 KByte/s Tx,  112098 KByte/s Rx.

Now, we are investigating cabling and routers, but I was wondering, is there any Solaris 10 configuration that might be throttling or otherwise impacting the performance of the second server (at the 1Kb packet size).

The machines are 1 hop distance from each other, directly connected through a router. Network statistics show no obvious errors, though if anyone has a good way to check let me know.

Answer 1

Have you checked that the speed and duplex are set correctly on server_one (and its switch)? Are server_one and server_two on the same network? Do they talk to server_three over the same interface? Do you experience the same performance issues between server_one and server_two?