Tudor, indeed, for your kind of point-queries, the Jun2010-SP2 release of MonetDB does a full table scan per query. In fact, all Feb2010 & Jun2010 releases do so. The upcoming Oct2010 release will build a (for now transient) hash-index on-the-fly with the first point-query (per column/attribute) that is then exploited with all subsequent point-queries (on that column/attribute) (for now, until the server is stopped/restarted). (This is what also Nov2009 and probably earlier releases did.) You can download a preview of the Oct2010 release candidate from our nightly builds at http://monetdb.cwi.nl/testing/projects/monetdb/Candidate/.DailyBuilds./ (you'd need to install from the "SourceSuperBall", though, as we do not provide any binary nightly builds, in particular no Dabian packages). Hope this helps you further. Regards, Stefan On Mon, Oct 18, 2010 at 11:45:10AM +0200, Tudor-Ioan Salomie wrote:

Hi,

I have recently tried doing some experiments with MonetDB, measuring its performance for the following dataset+workload: * 1 table (50 attributes), 1GB, comprising 3.000.000 tuples * the attributes are fixed length strings and integer types * the workload is query (SELECT) only, where each query has the form: SELECT FROM MyTable WHERE ; * the list of predicates is a conjunction of equality operators * more than 99% of the queries predicate on the same attribute (let's call it A1) that has a *very* good selectivity * the average number of predicated attributes per query is 8 * for this workload, averaging over 4000 queries, there is 1 result tuple per query (actual result tuples per query are in the range 0-10)

I have installed MonetDB on a 64-bit Debian machines, using the .deb files provided on the internet: * libmonetdb1_1.38.5-20100824_amd64.deb * libmonetdb5-server5_5.20.5-20100824_amd64.deb * libmonetdb5-sql2_2.38.5-20100824_amd64.deb * libmonetdb-client1_1.38.5-20100824_amd64.deb * monetdb5-server_5.20.5-20100824_amd64.deb * monetdb5-sql_2.38.5-20100824_amd64.deb * monetdb-client_1.38.5-20100824_amd64.deb

For running the queries I use the SQL front end. I have run the queries both through "mclient" (directly on the server) and through the JDBC driver (from another client machine). The MonetDB server is bound to run on only one core on the machine (though I have not seen any difference if running it on multiple cores).

To my surprise, the throughput for MonetDB for this workload is very low. I get about 3 queries per second, which seems to indicate that MonetDB does full table scanning only. Am I correct in this assumption? Is there anything that can be done to speed it up? Is this not a workload suitable for MonetDB?

Looking forward to your input!

Regards, Tudor.

------------------------------------------------------------------------------ Download new Adobe(R) Flash(R) Builder(TM) 4 The new Adobe(R) Flex(R) 4 and Flash(R) Builder(TM) 4 (formerly Flex(R) Builder(TM)) enable the development of rich applications that run across multiple browsers and platforms. Download your free trials today! http://p.sf.net/sfu/adobe-dev2dev _______________________________________________ MonetDB-users mailing list MonetDB-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-users

-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4199 |