On Mon, Nov 10, 2003 at 04:50:02AM -0500, Ken Corson wrote:
> Ed Tomlinson wrote:
> >On the otherhand, maybe we want to just eliminate QR and use something
> >like this.  Nodes have 100ms to reply with a Query Accepted otherwise the
> >query is deemed to have been rejected.   NG should be more that able to 
> 
> Ed, this is a feasible solution in certain environments. However, the
> optimal timeout window for every pair of nodes is going to vary widely.
> The round-trip (out and back) network + OS delivery time is likely to
> exceed 100ms for many routes. Trying to pick a one-size-fits-all value
> would _not_ be helpful for this project. QAs that are received too
> late are _at_least_ as wasteful as a QR (look at what happens in the
> meantime!). The nodes would need to measure the average round-trip
> time between them beforehand. This is very complicated to get right,
> and somehow I expect it would be counterproductive to attempt to
> implement it in this project, at this time. I posted another devl
> message earlier that discusses time handling in this kind of
> distributed environment ("a pointer on distributed programming").

We already have a timeout for Accepted. It's around 1 hop time - about 4
seconds IIRC. I don't think reducing it will be helpful.

>   I can't decide if a simple 'backoff for X seconds' message would be
> a decent gain over current situation or not. You have to account for
> the send time, receive time, time to act on the message, events that
> happen asynchronously during this time window ...

Make the QueryRejected's the backoff message. Don't specify a backoff,
just do randomized exponential backoff when you get a QueryRejected
message. And yes we need to work out some details - for example we don't
want a bunch of QRs coming in at the same time to trigger an exponential
backoff - so maybe we backoff only when the previous backoff has
expired, we have had 0 accepted and we get another QueryRejected.
-- 
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.

Attachment: signature.asc
Description: Digital signature

_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to