[netperf-talk] Testing on Freescale MPC8313ERDB

[Rick Jones]

Dominic Lemire wrote:
> Hello,
> 
> I need a quick opinion regarding the results below. Basically, I just 
> wanted to confirm that my "gigabit Ethernet" board (Freescale 
> MPC8313ERDB) is really capable of doing gigabit Ethernet... The bulk 
> data transfer test gives the same results when I connect it to a 
> computer with 10/100 Ethernet or gigabit, so I just want to check with 
> you before I complain to Freescale...
> 
> Thanks,
> 
> Dominic
> 
> ----- Connecting to a computer with 10/100 Ethernet ------
> PHY: e0024520:04 - Link is Up - 100/Full
> 
> # ./netperf -H 192.168.1.1
> TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.1.1 
> (192.168.1.1) port 0 AF_INET
> Recv   Send    Send
> Socket Socket  Message  Elapsed
> Size   Size    Size     Time     Throughput
> bytes  bytes   bytes    secs.    10^6bits/sec
> 
>  87380  16384  16384    10.01      94.08
> 
> ------ Connecting to a computer with Intel 82545GM Gigabit Ethernet (rev 
> 04) ------
> PHY: e0024520:04 - Link is Up - 1000/Full
> 
> # ./netperf -H 192.168.1.1
> TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.1.1 
> (192.168.1.1) port 0 AF_INET
> Recv   Send    Send
> Socket Socket  Message  Elapsed
> Size   Size    Size     Time     Throughput
> bytes  bytes   bytes    secs.    10^6bits/sec
> 
>  87380  16384  16384    10.00      93.48
> 


Before you fire-off a complaint to Freescale, a couple things to check:

1) while Linux (I am ass-u-me-ing you are using linux from the socket buffer 
sizes displayed) does autotune the socket buffer sizes, you might try explicit 
socket buffer size settings of something more like oh, 128KB:

netperf -H 192.168.1.1 -- -s 128K -S 128K

2) you should see what the CPU utilization happens to be during the test:

netperf -H 192.168.1.1 -c -- -s 128K -S 128K

and make certain your CPU is not saturated.  If there is any question whatsoever 
the remote CPU might be bottlenecking, you should check there too - add a "-C" 
after the "-c"

happy benchmarking,

rick jones