Cell Relay Archive

[nulldevice <nulldevice@eli.net>]

Adrian Rosoga wrote:

> Hi.
>
> I would like an answer to some basic questions about CAC.
>
> 1. Is there any dependency between the CAC and the hardware? Is there any
> generic CAC function that could be used to provide the answer if the
> connection can be accepted or not? Restated: when the hw change, the CAC
> change also?

The CAC is hardware specific.

> 2. If on some platform the switch fabric is changed (with a new device that
> has a different queuing and scheduling), is it necessary to change the CAC
> code?
> Is CAC dependent on the queuing structure?
> Do you think the CAC used on a certain ATM switch could be used with minor
> modifications on a Lucent switch (that has a quite different switch fabric,
> I suppose)?

Most modern switches implement minimal input queueing, a non-blocking
(oftenclocked at 2x the input rate) transfer matrix and output queueing. In this
case,
the output queueing is the major factor that would affect the parameters CAC
is trying to ensure. So the similarity between output queueing designs and
capacity
would determine the transportability of the CAC. In general, I would say "not
recommended"

> 3. Do you really see possible to write a "generic" CAC function that could
> be used irrespective of the actual hardware (I mean the code of the function
> remains the same when the hw is changed, just the input parameters of the
> routine change)?

This should be possible and is discussed in books and academic papers.

> 4. Is there any interest to try to write a "universal" CAC, good with any
> hardware? Is it possible?

In theory, the equipment manufacturer "guarantees" various qualities of
servicefor the traffic passing through their switches if their CAC is used.

> 5. Where is the best place for CAC to run (I mean at a ATM switch level)?

The CAC needs knowledge of the capacity already in use at each switch whenthe
call is proposed.