Cell Relay Archive

Cell Relay Retreat>List Archive>month:1997-Jan> msg00242



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Re: Lower bound for maximum ATM bandwidth?

  • From: Herbert Leitold <hleitold@iaik.tu-graz.ac.at>
  • Date: Wed, 29 Jan 1997 17:01:09 +0100

Steffen Jahn wrote:
[snip]
> 3. SONET OC3c operates at  3 * 9 * 90 * 8 * 8000/s = 155.520 mb/s
>    Available for ATM                                 149.760 mb/s
>    Available for ATM Adaption Layer            about 135.632 mb/s
>                                                      ============
> 
> The maximum throughput i ever got, was about 133 mb/s with UDP-over-ATM
> (mtu 9180, packet size 9140). But with all TCP/IP, LLC and AAL5 overhead,
> you can get theoretical about 134.5 mb/s. Anyone got so much with UDP or
> over an ATM API?

Your calculations are correct. 
On an unloaded switch, we could achieve with both UDP and TCP about 
133-134 Mbit/s.
But, reaching such maximas is -beside parameters such as socket-size, 
message-size, etc.- rather a question of computing power. 
E.g. when using LANE (MTU 1500) instead of ClassicalIP, ForeIP
(MTU9180/9188)
causing additional protocol stack processing, our TCP/UDP results using 
ULTRA 1's decreased from about 133 to about 95 Mb/s.

For a comparision of some different platforms you can have a 
look at our webpage,
   http://www.iaik.tu-graz.ac.at/RESEARCH/ATM/performance.html
or the "netperf submitted results" page via the Netperf page
   http://www.cup.hp.com/netperf/NetperfPage.html

Regards
   Herbert
-- 
Herbert Leitold (email: hleitold@iaik.tu-graz.ac.at)
IAIK Klosterwiesgasse 32/I  A-8010 GRAZ  Austria
URL.:  <http://www.iaik.tu-graz.ac.at/iaik.html>
Tel.: ++43 (316) 873 5516  Fax.: ++43 (316) 873 5520