For a graceful migration it might be necessary for the cache to shrink as the store fills... We do have online shrinking now, so that should be possible.
On Thu, Jul 06, 2006 at 02:07:23PM +0100, Matthew Toseland wrote: > Do you have any objections to me implementing the two level datastore in > the near future? To recap: > - We have a store and a cache for each keytype. > - The cache is 1/4 the size of the store. It caches everything that goes > through the node. I would have the existing store moved to the cache. > - The store only caches inserts, and only if it is the best node for the > insert so far on the chain. > - 1 in 100 successful requests, on each node, randomly turns into a > request at full HTL. Obviously the number is alchemy! 1 in 20 would be > too low as some requests will go 20 hops. > > Oskar's simulations show that this works much better than our current > scheme. -- 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/20060706/fe20f907/attachment.pgp>
