Degraded performance when using GRE over tinc

[Loic Dachary]

Hi,

In a setup where OpenVSwitch is used with GRE tunels on top of an interface provided by tinc, I'm experiencing significant performance degradation problems (from 100Mb/s down to 1Mb/s in the worst case) and I'm not sure how to fix this. The manifestation of the problem is, from the user point of view, iperf reports ~100Mb/s and rsync reports ~1Mb/s:

$ iperf -c 91.224.149.132
------------------------------------------------------------
Client connecting to 91.224.149.132, TCP port 5001
TCP window size: 23.5 KByte (default)
------------------------------------------------------------
[  3] local 10.0.2.27 port 57143 connected with 91.224.149.132 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec  99.4 MBytes  83.3 Mbits/sec


$ rsync --progress root at 91.224.149.132:precise-server-cloudimg-amd64-disk1.img .
precise-server-cloudimg-amd64-disk1.img
    32112640  12%   83.44kB/s    0:45:34

A tcpdump of iperf ( http://dachary.org/loic/good.cap ) shows no retransmission while a tcpdump of rsync  ( http://dachary.org/loic/bad.cap ) shows a frequent "TCP Dup ACK", "TCP Retransmission", "TCP Out-Of-Order". 

I'm not sure why this happens, any hint would be greatly appreciated :-)

Cheers

-- 
Loïc Dachary, Artisan Logiciel Libre

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 263 bytes
Desc: OpenPGP digital signature
URL: <http://www.tinc-vpn.org/pipermail/tinc/attachments/20140429/5a18db6f/attachment.sig>