Tinc over 3g problems?

[Guus Sliepen]

On Thu, Mar 26, 2009 at 03:44:11PM +0200, Thomas Arvanitis wrote:

> I am experiencing some weird problems in a setup with tinc where
> communication between the 'server' and the 'clients' occur over 3g
> connections.
[...]
> I tried to monitor packets arriving on Pegasus, but I wasn't able to
> make much sense out of it. Analyzing the traffic with wireshark shows
> a lot of packet re-ordering and retransmission though. And from what I
> show there were times that a tinc client was trying to establish the
> VPN with pegasus but that wasn't always successful!
[...]
> Is it possible that something related to the well known latency, rtt
> times and buffering issues of 3g, leads tinc to such a behavior?

It certainly is possible. The problem is that if you use TCPOnly and have a
lossy link, then both tinc's TCP connection and any TCP connection you tunnel
over it start retransmitting packets at approximately the same time after a
packet is dropped. This is degrading the performance. Also, the kernel will
buffer packets of TCP connections of both tinc and applications, causing very
long delays. The problem is described in more detail here:

http://sites.inka.de/~W1011/devel/tcp-tcp.html

Unfortunately you have at least two NATs in the way which prevents you from
switching to UDP, which is the only real solution to the problem.

-- 
Met vriendelijke groet / with kind regards,
     Guus Sliepen <guus at tinc-vpn.org>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: Digital signature
Url : http://www.tinc-vpn.org/pipermail/tinc/attachments/20090326/588d7cfa/attachment.pgp