[netperf-talk] netperf2.4.4 on linux, socket size issue ?

[Rick Jones]

>  > Also, at present, if one runs the omni tests over a 1GbE link, you will 
>  > notice that the autotuning takes the socket buffer size to the max, 
> 
> These omni tests are only present in the xml'ized netperf svn tip, not in
> 2.2, right? 

Nope, they are in the netperf2 top of trunk.

Here is a quick example of CSV output from an omni test behaving as 
TCP_STREAM with a sender on 2.6.24 and a receiver running something 
earlier.  This demonstrates a number of things - CSV output from an omni 
test, passing a filename to the csv potion to specify what to emit, and 
of course, the behaviour where it seems that the autotuning lets the 
socket buffer grow rather much larger than it needs to:

sut34:~/netperf2_trunk# HDR="-P 1"
sut34:~/netperf2_trunk# for i in -1 32K 64K 128K 256K 512K; do netperf 
-t omni -H oslowest $HDR -- -o foo -s $i -S $i -m 16K; HDR="-P 0"; done
OMNI TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to oslowest.raj 
(10.208.0.1) port 0 AF_INET
Throughput,Local Send Socket Size Requested,Local Send Socket Size 
Initial,Local Send Socket Size Final,Remote Recv Socket Size 
Requested,Remote Recv Socket Size Initial,Remote Recv Socket Size Final
941.32,-1,16384,4194304,-1,87380,4194304
925.15,32768,65536,65536,32768,65536,65536
940.97,65536,131072,131072,65536,131072,131072
941.00,131072,262144,262144,131072,253952,253952
941.36,262144,524288,524288,262144,253952,253952
941.36,524288,1048576,1048576,524288,253952,253952

The first result line there is letting the stack autotune.  The contents 
of foo are:

sut34:~/netperf2_trunk# cat foo
throughput,lss_size_req,lss_size,lss_size_end,rsr_size_req,rsr_size,rsr_size_end

The local system had rmem_max and wmem_max set to 4194304, the remote 
did not, so the above is also showing how autotuning is allowed to go 
farther than an explicit setsockopt() is.


> Would you accept a patch to have 2.2 spit out the ending
> size with a -v 2 in 2.2?

I think we have that covered in the omni tests.  And assuming the omni 
tests are received reasonably well, I plan on converting things like 
TCP_STREAM to use the omni code internally and then have its own 
"pre-omni" like output.  It would be to that that adding reporting when 
-v 2 is set would be added Ishould think.

rick jones