Re: [Pppext] RFC1990 - Sending multiple packet fragments

[James Carlson <james.d.carlson at sun.com>]

Dharanalakota, Divakar (Divakar) writes:
> Thanks for the detailed explanation. My aim is to do the interleaving of
> the voice packets along with the data packets and provide smaller
> serialization delay for the voice packets. I have some more questions
> from your explanation. 

For low-delay applications, I would suggest just not sending the
delay-sensitive packets through the MP part of the system.  Send them
with PPP Protocol Identifier 0021 (IP) rather than 003D (MP).  That
way, the peer won't be required to reassemble them in sequence.

> Can the sequence number have holes while sender is transmitting the
> packets. W.r.t the example, lets say 1B-a is having sequence number of
> 10, does 2BE need to have 13 only or can it start from some other number
> like 20. This means to say there is a jump in the sequence number from
> 13 to 20 while transmitting the packets. Here there is jump across the
> packets only, but sequence number is contiguous for all the fragments of
> a single packet. 

Such a scheme would only make things worse.

On the long term, what you suggest.  The microscopic result of doing
that, though, would be a short stall -- until the algorithm sees a
post-jump sequence number on all of the links, it cannot perform
delivery on those post-jump fragments.  It thus adds delay to the
link.

In other words, suppose you have those 5 links as before.  When you
transmit fragment number 20 on one of the links, the reassembler must
wait until all of the other four links also transmit fragments past 20
before that message can be delivered.  There's always the possibility
that one of the links that most recently sent 13 or lower might start
emitting fragments in the range 14 through 19.  There's no way for the
receiver to know that it won't.

See section 4.1 of the RFC; that's what the minimum sequence number
does.

I still think you need to explain what problem you're trying to solve.
You've got a mechanism mapped out (creative use of the sequence
numbers), but I don't see what problem that could address.

(If you go through the archives, you might find some earlier proposals
to use MP link bundling without ordering and without encapsulation,
particularly for very high speed links.  These didn't go very far.)

-- 
James Carlson, Solaris Networking              <james.d.carlson at sun.com>
Sun Microsystems / 35 Network Drive        71.232W   Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757   42.496N   Fax +1 781 442 1677
_______________________________________________
Pppext mailing list
Pppext at ietf.org
https://www.ietf.org/mailman/listinfo/pppext