Re: [Tech] Load balancing redux

Top Page
Author: Matthew Toseland
Date:  
To: tech
Subject: Re: [Tech] Load balancing redux
Delete this message
Reply to this message
gpg: Signature made Mon Apr 24 19:22:54 2006 UTC using DSA key ID 6554A22D
gpg: Good signature from "Matthew John Toseland (Toad) <toad@amphibian.dyndns.org>"
On Fri, Apr 21, 2006 at 10:16:13PM +0100, Michael Rogers wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Matthew Toseland wrote:
> > - We compute the overall request quota from the above.
> > - If there are n peers, and q quota, then each peer has a guaranteed q/n
> > requests. These can be sent regardless of what the other peers do.
> > - Some nodes will send fewer requests than their guaranteed share of the
> > quota, so after the basic quota shares have been allocated, some global
> > quota will remain. If all remaining requests above the basic quota fit
> > into this, we let all requests through.
> > - If not, we allocate the requests to the node which is the least above
> > its quota, followed by the one with next fewest extra requests...
> > until we run out of global quota.
>
> You've reinvented the bucket ;-)


:)

For anyone interested in this debate, most of it is happening on
devl@???.
>
> Cheers,
> Michael

-- 
Matthew J Toseland - toad@???
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.