Cell Relay Archive

[Paul Deneault <pdeneault@sympatico.ca>]

Thanks Walt.  Can you recommend a book that talks about this kind of stuff in detail
?

Regards,
Paul

Walt Weber wrote:

> Paul Deneault wrote:
>
> > Thanks for the response.  So let's see.  If I set a PCR=55000c/s and
> > SCR=45000c/s with MBS at 1200
> > cells over a cell rate of 46200 will be tagged with CLP=1 ?
>
> Close.  Assume no cells have been sent.  The first 1200 cells above SCR (and less
> than PCR) will be switched unchanged.  After that, cells that arrive at a rate >
> SCR < PCR will be tagged....but it's a little more complex than that.  Since the
> ratio of your PCR to SCR is about 5/4, about 1 cell out of 5 will be tagged,
> assuming you are sending at PCR.  At least, this is what I've seen from ATM
> switches.
>
> > Hmmm.  Ok here's
> > another question.
> > What should the CDVT be in this instance ? even if you'd have to guess ?  I
> > know it depends on the
> > interface.  Let's say a T3.  Would CDVT=20000 usec make sense ?
>
> 20000 is a large CDVT for a DS3, especially with a cell rate of about half a
> DS3.  It all depends on how 'clumped' your cells are.  If the traffic is shaped
> to PCR, you could probably get by with a CDVT of 250 or less.  Shaping makes you
> cells evenly spaced so CDVT isn't a big issue.  On the other hand if you don't
> shape, you need to find out how clumped you cells are to determine CDVT.
> Normally, it is calculated for the number of back-to-back cells in the cell
> stream.
>
> >
> >
> > Cheers,
> > Paul
> >
> > Walt Weber wrote:
> >
> > > Paul Deneault wrote:
> > >
> > > > Hello,
> > > >
> > > > Could somebody please explain the relationship between PCR, SCR and MBS
> > > > has it pertains
> > > > to a VBR channel ? Ok so I establish a VBR channel with the settings
> > > > PCR=55000 cells/sec
> > > > SCR=32000 cells/sec
> > > > MBS=1200 cells/sec
> > > >
> > > > Do these numbers make sense ?  what should they be ? is there a rule of
> > > > thumb ? Outside of the "suck it and see" approach, is there a forumale
> > > > to use ?
> > > >
> > > > Cheers,
> > > > Paul
> > >
> > > Don't know about rules of thumb, but....
> > >
> > > ATM switches in the network I work with police vbr as follows:  Cells at or
> > > below SCR are passed with no change.  Cells above PCR that exceed the CDVT
> > > are discarded.  Cells between PCR and SCR are passed with no change until
> > > the MBS value is exceeded, then these cells are tagged.
> > >
> > > BTW, the switches I am familiar with measure MBS in cells, not cells per
> > > second.
> > >
> > > I"m not familiar with routers, but it seems to me that if you are running
> > > IP over ATM AAL5 and want the all your traffic to end up in high priority
> > > ATM cells, you should have a MBS that matches with the size of the largest
> > > IP packet.  Then you can send IP packets at SCR that burst up to PCR.  In
> > > other words, each packet is sent at PCR, but the packets are spaced to keep
> > > the average rate at SCR.  On the other hand, if you want faster
> > > transmission and can live with tagging of ATM cells, you can set up the
> > > CDVT to allow a whole IP packet at the line rate.  You can then send at PCR
> > > with bursts at line rate.