[45925] in North American Network Operators' Group
Re: Satellite latency
daemon@ATHENA.MIT.EDU (Jim Mercer)
Thu Feb 28 12:27:11 2002
Date: Thu, 28 Feb 2002 12:27:23 -0500
From: Jim Mercer <jim@reptiles.org>
To: Eric Oosting <eoosting@sockeye.com>
Cc: nanog@merit.org
Message-ID: <20020228122723.N22359@reptiles.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Pine.GSO.4.44.0202281152590.24326-100000@fishy.sockeye.com>; from eoosting@sockeye.com on Thu, Feb 28, 2002 at 12:14:43PM -0500
Errors-To: owner-nanog-outgoing@merit.edu
On Thu, Feb 28, 2002 at 12:14:43PM -0500, Eric Oosting wrote:
> Is there something else specific to one-way satellite connections that you
> are attributing these problems with asymmetric routing to? There is
> nothing inherently wrong with asymmetric routing that causes high latency
> or perceived differences in usability of real-time connections like ssh.
> For instance, a large percentage of the traffic on the Internet is
> asymmetric. Anyone who is multihomed knows that half the time they send
> something to one upstream, it will come back a different upstream.
>
> Do you believe that there is something wrong with a downlink of 400ms and
> an up of 100ms? (guesses at satellite with dialup return connection one
> way latencies) I can't imagine that things would be better with a 400ms
> down and 400ms back up (as with a two-way sat connections) but maybe you
> can do different tricks with spoofing tcp acks to the traffic over the sat
> link when it is two-way.
in my experience, the upstream, typically via terrestrial, is more lossy
than the downlink.
this has nasty effects on ssh windowing. i haven't used telnet in a long
time, more specifically over sat links.
asymetric traffic over high-speed, low latency, low loss links is mostly
transparent.
when one path is clean, and the other is dirty, then the symptoms get worse.
also, asymetry is relative.
in one case, i had packets go high speed across north america (uunet/teleglobe/
sprint/alternet/etc/etc) over unencumbered fiber to the UK, then up via
satelite, down into the middle east into a terrestrial fiber network to the
dial-up port to the user.
the return path was dial-up over ancient infrastructure, to an overloaded
national backbone provider, via overloaded circuits to singtel, across the
pacific (using different networks depending on the phase of the moon), into
north america.
this is wholly different from have the outgoing packet traverse teleglobe
from newyork->los-angeles, and having the return packet go via alternet from
los-angeles->newyork. (although, depending on the phase of the moon, sometimes
teleglobe and/or alternet resemble the previous paragraph). 8^)
--
[ Jim Mercer jim@reptiles.org +1 416 410-5633 ]
[ I want to live forever, or die trying. ]
