Cell Relay Archive

[Jim Millar <jmillar@dl.com>]

Santosh S wrote:

> Hello,
>
> The IMA specs says that if there is an anomaly condition, Cells on
> that
> *LINK* are not sent to the ATM layer. So
>
> 1. we are losing ATM cells which might be perfectly alright, because
> of an
> *ANOMALY* ( Though one of the IMA objectives is that Cells should not
> be
> sent misordered to the ATM layer). Is it a good thing to do ?

I'm a little confused. How are you getting good cells on a link in an
anomoly condition?

By the way, if you get an anomoly condition on a link that has a lot of
cells buffered cuz of differential delay, it is perfectly acceptable to
still send these cells up to the ATM layer until the "delayed" point in
time that the anomoly was detected. Is this what you meant by losing ATM
cells that are perfectly alright?

> 2. Since the link is out of round-robin, effective incoming cell rate
> decreases. To handle this do we decrease the IDCR, or introduce Filler
>
> cells, on behalf of the link that is down, so that IDCR remains same ?

If your receive link state machine indicates the condition back to the
transmitter, then the transmitter will stop sending ATM layer cells on
the link. This will result in a lower IDCR at the receiver.

--Jim Millar