Re: Comments on draft-ietf-multi6-v4-multihoming-02

[Joe Touch <touch@isi.edu>]

Iljitsch van Beijnum wrote:
> On 11-nov-04, at 15:13, Joe Touch wrote:
>
> > Has anyone considered keeping the connection 'active' (i.e., 
> > keepalives) for sessions for around 1-2 typical RTTs of idle? That 
> > would catch TCP idles that still allow TCP bursts; anything longer 
> > than that and TCP is supposed to 'restart' its congestion control anyway.
>
>
> > That might be a happy medium to the 'all or none' keepalive. I.e.,
> > "keepalive for 1-2 RTTs of idle".
>
> Hm, wouldn't just one or two RTTs be a bit short? You may still get a 
> delayed ACK after 2 RTTs.

The delayed ACK should itself refresh things; I'm thinking of an 
otherwise completely idle interaction. The key here is that TCP expects 
things not to change for a few RTTs (pick a number). Designing things so 
that is not the case is likely to cause unintended impact on TCP, esp. 
in some very typical cases (web surfing over persistent connections).

> I think this basically boils down to the idea I was talking about 
> previously, where there are only reachability checks when there are 
> packets going in one direction. So when we transmit a packet, we start a 
> timer. When we receive a packet, we cancel the timer. When the timer 
> expires, we test reachability. Some optimization is possible with TCP so 
> that we don't start the timer for ACK packets with no data, but we do 
> stop the timer for those.

I'm suggesting that the above be tempered after a few RTTs. I.e., if the 
only traffic has been the reachability tests for more than X RTTs (pick 
X), then it might be OK to stop testing. This solves the "persistent 
chatter over idle channels" problem, and (appropriately?) allows 
underlying asspciations to be dropped/moved without substantial impact.

Joe

Attachment: signature.asc
Description: OpenPGP digital signature