[netperf-talk] Issues using parallel sessions of netperf

Brandeburg, Jesse jesse.brandeburg at intel.com
Wed Apr 13 14:40:26 PDT 2011


On Wed, 2011-04-13 at 14:06 -0700, Rick Jones wrote:
> On Wed, 2011-04-13 at 16:57 -0400, Srikanth Sundaresan wrote:
> > 
> > On 04/13/2011 04:54 PM, Rick Jones wrote:
> > > On Wed, 2011-04-13 at 16:31 -0400, Srikanth Sundaresan wrote:
> > >> Hi,
> > >> We are a group from Georgia Tech working on building a benchmarking
> > >> suite for home broadband internet. I've been testing netperf out as the
> > >> throughput tool.
> > >> In order to fully saturate the link, I'm running 3 netperf sessions
> > >> simultaneously. However, this causes 1 or two of the sessions to fail
> > >> with the following error:
> > >>
> > >> send_tcp_maerts: data recv error: Interrupted system call.
> > >> The output also has the following error: len was -1
> > >>
> > >> This only seems to occur on the TCP_MAERTS test; the TCP_STREAM test
> > >> seems to work fine (caveat: this is from a few dozen runs). Also, if I
> > >> run only 1 session, TCP_MAERTS works.
> > >>
> > >> Here's a sample output:
> > >> bin/netperf -l30 -P0 -fk -t TCP_MAERTS -p14536 143.215.129.122
> > >> bin/netperf -l30 -P0 -fk -t TCP_MAERTS -p14536 143.215.129.122
> > >> bin/netperf -l30 -P0 -fk -t TCP_MAERTS -p14536 143.215.129.122

I actually think your problem is because you're specifying port number,
there are some bugs in that area AFAIR.

I assume on the above that you forgot the -H option?


maybe try a different test to fill the link?
cd netperf-source
./configure --enable-burst
make -j
make install

this will do large receive bursts with 1 byte acks, much like normal
stream test.

[root at localhost ~]# netperf -t TCP_RR -l 10 -fk -H 10.23.23.181 -- -b10 -r1,64K
TCP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.23.23.181 (10.23.23.181) port 0 AF_INET : first burst 10
Local /Remote
Socket Size   Request  Resp.   Elapsed  
Send   Recv   Size     Size    Time     Throughput 
bytes  Bytes  bytes    bytes   secs.    10^3bits/sec   

16384  87380  1        65536   10.00    941194.07   
16384  87380 


> > >> send_tcp_maerts: data recv error: Interrupted system call
> > >>
> > >> In this example, one session failed. I tried opening 3 instances of
> > >> netserver on the server on 3 ports, but the same error occurred.
> > >>
> > >> I'm using netperf 2.4.5 on Linux 2.6.32. Please let me know if you need
> > >> more information. Any help would be greatly appreciated.
> > > 
> > > My initial guess is that the end-of-test timer is firing when it was not
> > > expected to - perhaps the rate of data flow through the broadband link
> > > into the netperf(s) is such that the remote's close does not happen
> > > before the watchdog timer pops?
> > > 
> > This testing is in an academic network (1Gbps connection between 2
> > servers). I was able to replicate it on a broadband link (though I only
> > ran the experiment a couple of times). Is there any more information
> > that I can give you that might help in debugging the issue?
> 
> How much time passes before the error message(s) appear?
> 
> happy benchmarking,
> 
> rick jones
> 
> _______________________________________________
> netperf-talk mailing list
> netperf-talk at netperf.org
> http://www.netperf.org/cgi-bin/mailman/listinfo/netperf-talk



More information about the netperf-talk mailing list