On Tue, Nov 11, 2003 at 01:22:21AM +0000, Toad wrote: > I will now merge the current 6323 (plus the recent improvement in RT > sorting) to stable. After that, we will see if the recent changes have > made much difference, and implement exponential backoff on > QueryRejected's, a strategy very similar to what we used during the > supposed golden age, which is simple to implement and has similar > effects to ian's proposal. Recent bugfixes, SFT, and other areas may > make a significant difference on their own. Anecdotal evidence suggests > this may be happening, so lets see what happens first.
Current status and battleplan: 1. 5032 contains major bugfixes and tweaks that when widely deployed should improve routing. 2. 5032 doesn't QueryReject based on bandwidth, unless the bandwidth is way over the low level limit. This should reduce the total number of QueryRejected's and start to kill off the positive feedback loop. 3. With less QueryRejecting, NGRouting should figure out when nodes are overloaded and not route to them. Of course this is a slightly longer term mechanism... 4. It will take a few days for 5032 to be widely deployed. 5. We may want to wait for this to test that status before implementing major new features etc. The only new feature w.r.t. load balancing that seems worthwhile is exponential backoff on QueryRejected's. This combines elements of various proposals and observations we have seen so far, and is simple and may just work. -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so.
signature.asc
Description: Digital signature
_______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
