[erlang-questions] gen_sctp throughput mystery

Anthony Shipman als@REDACTED
Thu Mar 8 22:07:24 CET 2012


On Fri, 9 Mar 2012 03:52:10 am Raimo Niskanen wrote:
> That is great news (that there is an explanation).
> 
> So the combination of default recbuf size of 1500 bytes in gen_sctp,
> the (Linux ?) kernel default net.sctp.sack_timeout = 200 (RFC recommended),
> and sending packets larger than recbuf size choaks the throughput to
> 1000/net.sctp.scack_timeout packets per second...
> 
> What a gotcha!
> 
In inet_int.hrl in R15B, the default for recbuf is 1024, while sndbuf is 
65536. When I set recbuf to 65536 in my test server the delays disappear. I 
can get up to 50MB/s from client to server before the eagain problem hits.

-- 
Anthony Shipman                    Mamas don't let your babies 
als@REDACTED                   grow up to be outsourced.



More information about the erlang-questions mailing list