From: ccie2be (ccie2be@nyc.rr.com)
Date: Fri Jul 30 2004 - 12:29:48 GMT-3
Hi Daniel,
Thanks for that very detailed and insightful explanation. It is truly very
appreciated and helpful.
Do you recall seeing a post yesterday from Richard Dumoulin. He suggested
creating 4 telnet sessions and pinging simultaneously from all 4 telnet
sessions to generate a greater traffic load.
I asked him how exactly to do that but didn't hear back. Have any ideas
about how I could do that?
Thanks again. Tim
----- Original Message -----
From: "Daniel Ginsburg" <dginsburg@mail.ru>
To: "Group Study" <ccielab@groupstudy.com>
Sent: Friday, July 30, 2004 10:11 AM
Subject: Re: Generating enough pings to make dialer load threshold kick in
> As far as I know timeout parameter affects only how long router waits
> for reply before printing '.' instead of '!'.
>
> I ran ping with default timeout, with 0 timeout and with timeout 100.
> All three completed in approximately same time. My conclusion is that
> rate of packets dictated only by round-trip time.
>
> Please note that 50% is upper bound. It can be significantly less is
> case of LFN (long fat pipe).
>
>
> On Fri, Jul 30, 2004 at 03:56:58PM +0200, Richard Gallagher wrote:
> > What about setting the timeout to zero? Then we don't ever hang around
> > and wait for a response. We just send as fast as the router can.
> >
> > On Fri, 2004-07-30 at 15:29, Daniel Ginsburg wrote:
> > > On Thu, Jul 29, 2004 at 10:50:38AM -0400, ccie2be wrote:
> > > > Hey Richard,
> > > >
> > > > I don't dispute or disagree with what you're saying but how do know
that 4
> > > > simultaneous pings with a packet size of 1500 will load the channel
to over
> > > > 80%? How do you know what load exactly that will put on the
channel? If
> > > > you don't know exactly what load that puts on the channel, how do
you know
> > > > that that is NOT, for example, a 65% load or 75 % load?
> > > >
> > >
> > > Unlike many other ping implementations which send 1 echo request per
> > > interval cisco's one sends next echo request as soon it receives echo
> > > reply or waits for timeout if request or reply is lost. So average
> > > bandwidth utilization in one direction with one 'ping a.b.c.d size X'
> > > will never exceed 50%.
> > >
> > > Let me illustrate this with the diagram
> > >
> > > ---
> > > ^ BW
> > > |
> > > | (1) (3) (5)
> > > |----- ----- -----
> > > |
> > > | (2) (4)
> > > ----------------------------------------->
> > > Time
> > >
> > > (1) router transmits ping request
> > > (2) router waits for ping reply
> > > (3) router transmits next ping request
> > > (4) router waits for next ping reply
> > > etc.
> > >
> > > So router uses the link almost[1] exactly half of the time. Please
note
> > > that this 50% figure almost[1] doesn't depend on size of echo
> > > request/reply.
> > >
> > > [1] I'm saying almost because router needs to ponder a very short
period
> > > of time before replying to echo request. This period of time may be
> > > negligible or not depending on speed of the link.
> > >
> > > Two simultaneous pings will theoreticaly saturate the link. Run four
to
> > > make sure ;)
> >
> > _______________________________________________________________________
> > Please help support GroupStudy by purchasing your study materials from:
> > http://shop.groupstudy.com
> >
> > Subscription information may be found at:
> > http://www.groupstudy.com/list/CCIELab.html
> >
>
> --
> dg
>
> _______________________________________________________________________
> Please help support GroupStudy by purchasing your study materials from:
> http://shop.groupstudy.com
>
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Sun Aug 01 2004 - 10:12:07 GMT-3