Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [RPRWG] RAH: Re: Minutes of Rate Ad Hoc meeting




Offer,

I believe we are largely in agreement. A few comments on some of your
points:

>I believe that all these behavioral parameters should be implemented in
>the ingress points to the ring via policers/shapers and not by relying
>on degree of loss on the wire itself.

In general, I agree. However, I could imagine some cases where it would be
advantageous to not do so for traffic that wants to be as greedy as
possible, and doesn't care about some packet loss (such as some UDP based
services).

>That is how can one
>factor in the possibility of losing a packet in transit in some very low
>probability and under very specific condition, in the overall parameters
>of admittance to the ring?

Same as how it's done in IP networks.

>Furthermore, if a user traffic is rejected at the ingress point, the
>user gets an immediate feedback. Assuming such a feedback is useful, how
>would a network do it if a given packet was dropped not at the ingress
>but rather somewhere on the ring, say on the 5th node in a 10-node path.

Agreed. If one has to drop packets, its always best to do so as early as
possible (in both time and distance from sender). However, it must always be
possible to deal with packet drops at any point in time or distance.

>Again, I hope that I've managed to clarify my views, but I am not sure
>if we "violently agree" or still have some differences.

I think we very largely agree. The above points are about what is possible,
not what is needed. I believe that this is likely to be just a fun mental
exercise. To know for sure, we need to examine whether allowing packet loss
in extreme cases actually buys enough in terms of delay/jitter for other
services or utilization for the ring as a whole to make it worth
considering. Until and unless someone can show this, we should not consider
allowing it.