Re: WG Review: Low Extra Delay Background Transport (ledbat)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

On 2008-10-31, at 10:45, ext Robert Elz wrote:
This looks like useful work to do, and to me, the charter mostly
looks fine, just one point.

The (proposed) charter says ...

 | * operate well in networks with FIFO queueing with drop-tail
 | discipline

which in itself is OK, but doesn't say anything about links that
don't have those characteristics.

The charter actually talks about non-FIFO-droptail queues. Below is Bruce Davie's summary, which sums up the intent very nicely:

On 2008-10-24, at 23:35, ext Bruce Davie wrote:
* operate well in networks with FIFO queueing with drop-tail discipline, * where available, use explicit congestion notification (ECN), active queue management (AQM), and/or end-to-end differentiated services (DiffServ).

The first bullet says "deal with the world as it is"; the second says "deal with the world as you wish it were"

I think that is a very sensible approach.


For different queueing strategies, I doubt there'd be a problem, but
I might expect the charter to be explicit that for (what I will call here)
intelligent queueing (almost anything other than FIFO that's useful)
the resulting protocol from this group will at the very least be
easily distinguishable (without necessarily resorting to DiffServ
mechanisms) from regular traffic.

It could be easily distinguishable (for example, we could decide to use the LTBE DSCP for all packets), but it doesn't need to be. LEDBAT is doing an end-to-end algorithm that is intended to work without any router support. (If there is router support, it'd obviously be nice if it performed better in some way.)

 Of course, is the result is an
IP application protocol that is not TCP, that's not going to be
an issue, but if it is a different behaviour of TCP stacks for
background traffic, it might not be so easy for intermediate routers
to be able to separate the background traffic from other traffic.

Perhaps more difficult, I'm not sure, would be how well the protocol is
to function with random drop, or RED, instead of drop tail.  I'd hope
it would be "at least as well as drop-tail", but I'm not sure, and
the charter doesn't say.

Right, the intent is that the LEDBAT mechanism should work at least as well if some sort of network help is available than when there isn't. Maybe we should make that explicit.

Thanks for your comments,
Lars

_______________________________________________

Ietf@xxxxxxxx
https://www.ietf.org/mailman/listinfo/ietf

[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]