Cell Relay Archive

[Paul Koning <pkoning@lucent.com>]

Baris Aksoy wrote:
> 
> > > As far as I know, it is used in AAL for detecting:
> > > 1. Lost cell
> 
> > > 2. Mis-inserted cells.
> 
> I'd agree with Thomas. Suppose one of the switch in the network is
> malfunctioning in a way that, it screws up during switching process. I mean,
> when queueing and sequencing, it may mix up the the cells that belong to same
> connection so that cell#1 is scheduled much later than cell#2. The switches
> on the way to destination doesn't process AAL layer, so it doesn't know
> sequence numbers. You have a disorder at the receiving end.
> 
> Based on today's switch architectures, this shouldn't happen. Have you ever
> experienced something like that on the field?

It shouldn't happen no matter what the switch architecture.  In other
words, it's a hard requirement that the switch does not do this.

Of course, hardware can break, and guarantees can fail to be met.

It makes no sense at all for anyone to consider implementing a 
cell order restoring function.  The most that would make sense is an
order
checker.

Then again, given the tiny sequence numbers, the probability that a
misinserted cell is *not* detected is quite high.  Realistically,
only AAL5 has a reasonable chance of dectecting such things thanks
to the CRC-32.

	paul