salo at networkcs.com
Thu Nov 6 19:09:00 UTC 1997
> Date: Thu, 6 Nov 1997 10:05:31 -0500 (EST)
> From: "Paul D. Robertson" <root at gannett.com>
> Subject: Re: New MAE-EAST
> I'm sure that's a part of it, I initially saw a lot of dropped packets
> through a couple of ATM clouds. I'm seeing some improvements in
> some of the providers, however, given the trumpeting of ATM (magic bullet
> syndrome), it seems that it's just not something which happens correctly
> by default.
Some ATM switches are better than others. In particular, some of the
early ATM switches had very small buffers. They simply weren't designed
to support bursty data like IP.
There are a number of "good" ATM switches, at least in the sense that
they have adequate buffering, available.
Depending on whether you are using ATM as a fine-grained, relatively
static multiplexing service (CBR) or as a bandwidth-on-demand service
(UBR or ABR), the amount of buffering in the network may make a difference.
This is probably a good question to ask your prospective service provider.
> before we go off on the 'nothing happens correctly by
> default' tangent, it's just been my general observation that whenever
> my packets have been transited over ATM, my latency has been less than
> ideal. I would have figured that oversubscription would result more in
> lost packets and timed out connections (which were also seen, but more
> easily screamed about) than latency, but I guess that's a factor of how
> oversubscribed the line is.
While I don't know your particular circumstances, lots of things can cause
latency. Distance, or perhaps unexpected distance, is often an explanation
for propagation delay. Some carriers may backhaul your ATM traffic to a
switch far away. So, you don't really know what latency to expect unless
you know the physical path your data traverse.
By the way, the same sort of thing can happen can happen with point-to-
point leased lines. We recently had a line which used a rather
circuitous route to reach its destination. We asked the carrier to
find a shorter path, which they did.
More information about the NANOG