Long-haul problems - connections stuck in slow start

[Peter Wemm]

On 1/4/14, 10:38 AM, Peter Wemm wrote:
> We're seeing some unfortunate misbehavior with tcp over an intercontinental
> link.

I forgot to mention, the socket buffers have been generously tuned for these
endpoints. They don't seem to be being used while the sender is in some sort
of limited transmit mode.

When reading the tcpdumps, keep in mind the window scaling factor is 11 and
that's wasn't captured in the session.  You'll have to manually compensate
when reading the mid-session dumps.

XXX footnote:  It seems turning SACK off makes a huge difference for this
connection.  The server must have been running with sack disabled when it
was working and hadn't been saved in sysctl.conf. Turning sack off on the
server again has raised the throughput from 8K/sec to 32MB/sec.  That's a
nice 400x speedup.

I'm investigating.

-- 
Peter Wemm - peter at wemm.org; peter at FreeBSD.org; peter at yahoo-inc.com; KI6FJV
UTF-8: for when a ' just won\342\200\231t do.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 246 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/freebsd-net/attachments/20140104/1008c445/attachment.sig>