On Wed, Nov 26, 2003 at 01:00:26AM -0500, Ken Corson wrote:
> Toad wrote:
> >Hypothetical:
> >Routing works, so we have a 20% success ratio.
> >The average filesize is 200kB (this is about right on the current
> >network, check your datastore - but maybe we need to gather more
> >accurate stats on it).
> >We have a 256kbps uplink i.e. 32kB/sec, of which we can use all (this is
> >optimistic).
> >We get a mere 10kqph incoming, and accept all of it.
> >
> >I will now demonstrate that this is impossible:
> >10kqph * 0.2 = 2kqph.
> >2000 * 200kB = 409,600,000 bytes
> >409,600,000 bytes / 3600 seconds = 113,777 bytes per second, for
> >trailers alone, assuming no connection and search overhead.
> >
> >Ideas?
> 
> this is based in pure reality :) as a practical matter, the times at
> which trailers will be created obeys a Poisson distribution. This
> is basic queueing theory in a nutshell. It is becoming clear that
> a certain bandwidth capacity will determine an optimal qph for any
> given node. We cannot control the success rate of queries, however,
> it should average out over some period of time. We seem to have
> acknowledged that we cannot accept as many queries once we become
> burdened (or blessed really) with "too many" trailers. This is the
> balance we seek between query traffic and trailer traffic. We can
> prioritize between trailers by dedicating more, or less, bandwidth
> to individual transfers. The goal is to service each one "fairly,"
> while attempting to maintain a relatively constant number of active
> trailers over time.
> 
> My hat is off to Toad for identifying hard limits that we simply will
> never go beyond! It frames the overall problem quite concisely. I
> suspect these variables could also help to produce a more meaningful
> or accurate %load factor.
> 
> There is an obvious inverse relationship between (average) individual
> trailer transfer rates and the number of active trailers... many slow
> or few that are fast - take your choice. But the first step is to
> approximate the optimum qph for a node with a given bandwidth.

Can we even do that?
We don't know the success ratio, which may well be a function of load
anyway!
> 
> -Ken
-- 
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