On Sun, Feb 13, 2011 at 12:27 AM, Marcus Grimm <mgr...@medcom-online.de>wrote:
> So my brain isn't that lasy, we count 3 syncs.... :-) > > When you turn journaling to off you will most likely > see less syncs, probably 2 in your case. So that is all > in line.. > > Marcus > > Marcus, you're right about general journaling, I also found 3 flushes in the Process Monitor log and the number of commits per sec is about 12-15. And looks like journal_mode=off requires only one flush. Interesting that WAL requires also only one sync (single flush in the log) and the speed about 40-50 commits confirms this. So there's also another reason to use WAL. Although the page http://www.sqlite.org/draft/wal.htmlalready says "4. WAL uses many fewer fsync() operations...", it attributes more to the consistency issues, but this tests also contributes to "1. WAL is significantly faster in most scenarios....". So the hint about wal can look like this: if possible number of commits per second can vary up to 30 per second, one should use WAL. Max Vlasov _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
