From: Ilya Mazhara (willy@xxxxxxxxxxxxxxxx)
Date: Wed Jan 17 2001 - 03:42:49 GMT-3
CAR is incoming option (provider point of view) cause its for dropping
GTS is outgoing option (user view) cuase its for shaping
Robert DeVito wrote:
>
> What are key words to look for in determining to use CAR or GTS? Thoughts?
>
> Thank you,
> Robert
>
> ----Original Message Follows----
> From: Brian Hescock <bhescock@cisco.com>
> Reply-To: Brian Hescock <bhescock@cisco.com>
> To: Roger Dellaca <rdellaca@bpopca.com>
> CC: jconnary@cisco.com, ccielab@groupstudy.com
> Subject: RE: traffic shaping on a serial line
> Date: Tue, 16 Jan 2001 21:37:11 -0500 (EST)
>
> Don't forget about Generic Traffic Shaping (GTS), which is always an
> option.
>
> Brian
>
> On Tue, 16 Jan 2001, Roger Dellaca wrote:
>
> > as far as CAR, I think it's a good solution, but the lab was written in
> Sep 99 (se the "topology" page), was CAR even around? The best solution now
> may not be the answer at that time.
> >
> > I can see the logic of wanting to put the traffic-shape on both ends. In
> the answer given, the access-l has telnet as the destination, and it's on
> the router with the referenced token-ring, so the packets are sourced from
> the token ring with destinations of telnet daemons somewhere else in the
> network, as you state.
> >
> > So the logic I would attach to this is that it will slow down the input
> to telnet if you cut & paste large volumes of commands! - but has no effect
> on the speed with which you receive back screens.
> >
> > >>> "Connary, Julie Ann" <jconnary@cisco.com> 01/10 2:17 PM >>>
> > That's what I thought of doing at first - and then the lab answer was
> > to use traffic -rate shaping on the serial so I questioned why only on
> > one side.
> >
> > This is the fatkid lab on performance and queing at fatkid.com
> >
> > Also - does token-ring support CEF? The docs say you have to have CEF to
> do
> > CAR.
> >
> > Julie ann
> >
> > At 08:04 PM 1/10/2001 +0200, you wrote:
> > >Ann,
> > >
> > >What about using input and output CAR on R3?
> > >This way shouldn't you force telnet (TCP) to back-off due to the
> windowing
> > >mechanism?
> > >Objections?
> > >
> > >-----Original Message-----
> > >From: Connary, Julie Ann [mailto:jconnary@cisco.com]
> > >Sent: Wednesday, January 10, 2001 3:53 PM
> > >To: ccielab@groupstudy.com
> > >Subject: traffic shaping on a serial line
> > >
> > >
> > >Hi,
> > >
> > >if you have the following:
> > >
> > >
> > >
> > >R1 --------serial at 64K ---------------R3 --------Token Ring
> > >
> > >You want to limit the users on the token ring from telnet using over 32K
> of
> > >the bandwidth,
> > >
> > >So I know on R3 you would use GTS to limit telnet to 32K.But what about
> the
> > >reverse traffic coming from R1?
> > >
> > >Don't you have to traffic shape there too? The fatkid labs only shape on
> > >one side of the link - the R3 side.
> > >Seems to me that in telnet you issue a command and get a screenfull of
> data
> > >- so more traffic would
> > >be coming back and you would want to also limit on the R1 side.
> Thoughts?
> > >
> > >I guess you would assume that the telnets are ONLY sourced from the
> > >TokenRing - so
> > >your access-list on R1 would have to be:
> > >
> > >access-list 101 permit tcp any eq telnet any
> > >
> > >vs. the access-list on R3:
> > >
> > >access-list 101 permit tcp any any eq telnet.
> > >
> > >
> > >Julie Ann
> > >
> > >------------------------------------------------------------------------
> > > Julie Ann Connary
> > > | | Network Consulting Engineer
> > > ||| ||| Federal Support Program
> > > .|||||. .|||||. 13635 Dulles Technology
> Drive,
> > >Herndon VA 20171
> > > .:|||||||||:.:|||||||||:. Pager: 1-888-642-0551
> > > c i s c o S y s t e m s Email: jconnary@cisco.com
> > >
> > >------------------------------------------------------------------------
> > >
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:27:32 GMT-3