BTW, the experience with dedup is that doing things off-line means never catching up, while doing them online means going slow.
You might cache as much as you can in memory then go slow when you miss the cache... In practice I think it's best to separate data and metadata devices so that you can make metadata as fast as you can, writing COW-style, like ZFS does, and caching in memory on clients and servers as much as possible for all data writes during the writes until they are committed. Nico -- _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users