From: chet johnson <chejohns@ix.netcom.com>
Subject: Re: Simulating network latency in a testbed
Date: 1997/02/02
Message-ID: <32F53B14.59E@ix.netcom.com>#1/1
references: <32F21879.6E6F@vines.etn.com> <32F2601A.2D8E@mindspring.com>
content-type: text/plain; charset=us-ascii
organization: Netcom
x-netcom-date: Sun Feb 02  7:09:29 PM CST 1997
mime-version: 1.0
reply-to: chejohns@ix.netcom.com
newsgroups: comp.dcom.net-analysis,comp.dcom.frame-relay,comp.protocols.tcp-ip,comp.databases.oracle.misc
x-mailer: Mozilla 3.01Gold (Win95; I)



Jeff,

I agree with Lawrence. I use a link simulator in my lab, but have found
it most suitable for demonstrating to the developer how the application
will "feel" to the desktop. By performing a protocol analysis and
evaluating the results, you can also look for specific remedies for an
application's sensitivity to latency.

The tool I'm looking for is a defensible measure of network latency.
Today I use ping (varying sizes). Since, most of the time, this is a low
priority protocol my results are easily challenged. Ideally, I'd love to
develop or find a mathmatical model that I can use and defend.

chet johnson
chejohns@ix.netcom.com

Lawrence L. Baldwin wrote:
> 
> Jeff Paige wrote:
> >
> > I am trying to do an analysis of network latency requirements for a
> > set of Oracle-based client/server applications which will run over
> > national and international WAN circuits (mostly frame-relay).  Since we
> > Jeff Paige
> > Eaton Corporation
> What you probably want is a link simulator...try searching on that.
> 
> I've worked with T1 and T3 simulators that allow you to vary link speed
> and link latency (in either direction).  Unfortunately, it's been a
> while and I can't remember any vendor names.
> 
> Another option is doing a packet level transaction benchmark in a LAN
> environment (this is the approach we use).  Basically, you capture each
> transaction using a standard LAN protocol analyzer.  Then you take the
> protocol trace and isolate how much of the delay is due to LAN latency
> and speed.  Finally, you can extrapolate these results using typically
> WAN link speeds, link propagation delays, and estimated congestion
> delays.
> 
> --
> -----
> Lawrence Baldwin                                baldwinl@mindspring.com
> System Management Technologies, Inc.
> --Internet Protocol Analysis Services
> web: http://www.netkb.com                       voice: 770-226-0590
 

