I'm not so sure how fast SQLite4 will be with respect to full scans.
IMHO row stores have an advantage when scanning over multiple columns.
Concerning dropping the index and recreating it on the other side. Its
doable but very bad from a performance point of view.
If you know how the other side will access the data, then the best
option is to build in parallel the indexes over all the data chunks (on
every cluster node), and then send the indexed chunks on the other side.
Having doubled data (table+full covering index) in these chunks, halves
the I/O bandwidth of the whole cluster.
Also i should point that our main use case is OLAP processing and not OLTP.
lefteris.
On 05/03/13 10:51, Nico Williams wrote:
SQLite4 gets this right... Of course, it's not been released.
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
