600ms

From: Jon Mansey ([email protected])
Date: Mon Nov 10 1997 - 18:36:10 EST


List,

I think I have discovered the source of my confusion. The 600ms used in the
throughput calculations applies to duplex satellite connections. With 250ms
for each up-down trip, plus some extra for processing, you get to 600ms.

My company provide simplex IP over satellite, the return side being mostly
terrestrial, and that is where my confusion came from, as I regularly see
ping times in the region of 300-400ms.

So now,

>E.g. I think Microsoft adopts 8kByte window => max thruput = 8192 x 8 / 0.6s
>= 109 kbps.

becomes 8192 x 8/0.35 = 187.2 Kbps. This is more like what I see over my
connections, the region of 20-25 KBytes/sec.

Just a footnote to all this, it should be remembered that this is "per
socket". With multiple sockets opening over a E1 link, each user sees
similar throuput, up to the capacity of the link.

It would be interesting to see a list of various TCP stacks with their
window sizes. Anyone?

Jon.

[email protected] Chief Science Officer
------------------------------------------------------------------
InterPacket Group, Inc. http://www.interpacket.net
501 Santa Monica Boulevard
Suite 702 (310) 656-8333
Santa Monica, California 90401 (310) 656-8326 fax
------------------------------------------------------------------



This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:32 EST