torej
asked on
Very low TCP/IP performance
One Ultra2/2x200 is connected to a Ultra1/167 via a 3COM 100Mb switch.
Both machines are running Solaris 2.5.1, including all recommended patches.
On the Ultra2 I am running ARCserve/OPEN for backup.
This software also backs up the Ultra1 over the network.
The problem is extremely low performance. I am not getting anything more than 3-4 MB/minute.
The switch is running in full duplex mode, the Ultras are using SunSwift, and I have patched wit the latest hme-patch (104212-05) which should enable the SunSwift cards to run in full duplex.
Please give me help on this.
Both machines are running Solaris 2.5.1, including all recommended patches.
On the Ultra2 I am running ARCserve/OPEN for backup.
This software also backs up the Ultra1 over the network.
The problem is extremely low performance. I am not getting anything more than 3-4 MB/minute.
The switch is running in full duplex mode, the Ultras are using SunSwift, and I have patched wit the latest hme-patch (104212-05) which should enable the SunSwift cards to run in full duplex.
Please give me help on this.
Better if you isolate the problem of ultra's or 3com switch.
Make one cross cable and connect both ultra machine directly
without the switch and see the performace, if this is okay
then problem lies in switch. Now call up 3com, you nay want to
check PACE feature of 3com switches.
vinay
Make one cross cable and connect both ultra machine directly
without the switch and see the performace, if this is okay
then problem lies in switch. Now call up 3com, you nay want to
check PACE feature of 3com switches.
vinay
I got similar performances when running TCP over ATM. The reason was that all TCP parameters where set by default for connections over small networks. Here is a script that I used to set the parameters which made it go up to 100Mb/s:
#!/bin/sh
ndd -set /dev/tcp tcp_xmit_hiwat 65535
ndd -set /dev/tcp tcp_recv_hiwat 65535
ndd -set /dev/tcp tcp_mss_def 9180
ndd -set /dev/tcp tcp_sth_rcv_hiwat 128000
ndd -set /dev/tcp tcp_rwin_credit_pct 100
ndd -set /dev/tcp tcp_cwnd_max 65535
ndd -set /dev/tcp tcp_naglim_def 2048
I'm not sure what the mss is on fast ethernet. Correct the line that says tcp_mss_def 9180 accordingly
hope this helps,
Philippe
I'll be glad to resubmit it as answer if it does solve your problem
It's definately a result of something the switch does. I have
the same problem (2 Ultra's with bayswitches in between). Removing the switches from the picture 'solved' the problem.
Tweaking the parameters like Philippe suggests helps, but I'm
still looking for the ideal parameters.
the same problem (2 Ultra's with bayswitches in between). Removing the switches from the picture 'solved' the problem.
Tweaking the parameters like Philippe suggests helps, but I'm
still looking for the ideal parameters.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
are running
together with other UNIXs in the same net segment.
Don't know if this also applies if they are connected via a switch (ask Sun).