On Oct 21, 2010, at 10:05, Pavel Ivanov wrote:
> I think it's not related to fragmentation, but to fill percentage of
> b-tree pages. I guess your reconstructed table is much less in total
> size than your initial one. Also does changing cache_size changes
> above numbers?
Interesting. We have been using sqlite3_analyze on some tests and
finding that we can get our dbs very fragmented. It doesn't report fill size
as far as I can tell.
We'll play around with cache_size to see if that does anything useful
for us in the meantime.
> What size do these tables have?
About 2.2GB with about 4 million rows.
> What bottleneck appears to be in 3-hour query execution? Is it disk thrashing?
Yes.
I've tried different strategies in the past. Vacuum and the rebuild
both seem to help quite a bit. I don't understand the file layout all that
well right now, so I don't completely understand how the index is traversed.
--
Dustin Sallings
_______________________________________________
sqlite-users mailing list
[email protected]
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
