On Tue, Aug 22, 2006 at 11:07:44PM +0100, Matthew Toseland wrote: > On Tue, Aug 22, 2006 at 09:40:03PM +0200, David 'Bombe' Roden wrote: > > On Tuesday 22 August 2006 21:20, Matthew Toseland wrote: > > > > > > Couldn't we simply measure the time it takes us to process an > > > > incoming packet, i.e. the time spent between receiving of the > > > > packet and having sent out the response packet? > > > That only measures CPU usage. We want to detect network overload too. > > > > The time till a packet is ready to be sent is CPU load, the time it > > takes till the packet was finally sent measures the network load. > > What about packet loss? I suppose we're pretty screwed anyway on really > lossy wifi links, and the best thing is to layer something on top, then > layer freenet on top of that?
Well, I suppose that's where link level congestion control comes in. Thing is, this doesn't affect the rate at which we can receive *requests*. Although it does affect the time it takes for them to complete, so the RTT goes up, so maybe we're okay. I'll implement the measurement. :) > > > > David -- Matthew J Toseland - toad at amphibian.dyndns.org Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/tech/attachments/20060822/e8cb72dd/attachment.pgp>
