Cell Relay Archive

Cell Relay Retreat>List Archive>month:1995-Apr> msg00341



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

Re: TCP window size? Cell-level pacing?

  • From: kjh0@gte.com
  • Date: Fri, 21 Apr 1995 20:15:25 -0500, Fri, 21 Apr 95 20:41:55 PDT

In article <3n969h$1mh@prop.wes.army.mil>, <sweberg@wes.army.mil> writes:

> 
> My next question is about cell-level pacing.  I read a paper
> (TCP/ATM Experiences in the MAGIC Testbed) that suggested that cell-
> level pacing dramatically improves throughput over an ATM WAN.
> What is cell-level pacing?  How do you go about implementing it?
> Does it apply at the unix socket level, or at the ATM level?
> 
Cell-level pacing refers to rate control. If you have a workstation
pumping at OC-3 rates thru a switch to a DS-3 connection on the wide-area
you are going to fill up buffers in the switch. Even if your average rate 
is well below DS-3 your bursts may be a OC-3 line rate which can fill
up the average switch buffer and cells will be discarded causing TCP 
retransmitts (a very bad thing).  Most of the ATM NIC's allow you to 
manually configure rate control such that the outgoing data is paced and
hence will be kinder to those light to medium buffered switches.

> Thanks for any and all help, suggestions, and pointers to info!
> 
> Scott D. Weberg
> Computer Scientist
> U.S. Army Corps of Engineers

Ken