Re: TCP (and SCTP) sucks on high speed networks

[<sm@mirapoint.com>]

---- Original message ----
>Consider a 10Gbs link to a destination half way around the world.  A packet
>drop due to link errors (not congestion or infrastructure products) can be
>expected about every 20 seconds.  However, with a RTT of 100ms (not even
>across the continent), if a TCP connection is operating at 10Gbs, the packet
>drop (due to link error) will drop the rate to 5Gbs.  It will take 4 *MINUTES*
>for TCP to ramp back up to 10Gbps.

I suppose you have some particular real life example of this scenario
but nevertheless it's a fairly unusual one and the fact that TCP does not
address its needs isn't surprising.  Do you just have one connection
on this 10Gbs link?  Maybe you should consider an alternative protocol,
such as STP:
 
        http://www.hippi.org/cST.html

Cheers,
-- Sam
 
P.S. As an FYI, at least one OS I am aware of (IRIX) has a systuneable that
will make TCP violate the RFCs and behave extremely aggressively for
high speed, low latency links, to address this kind of need.  I guess
that doesn't really help your long-distance case though.
------------------------
Sam Manthorpe, Mirapoint