[MonetDB-users] Table column order and type
Hi, From what I understand indexes are indicative[1], sortedness is determined at runtime[2] and the storage model should not effect the general schema design[3]. Has anyone had any experience with large time series where the column order and types have effected performance? For example having data with less variance and "uniqueness" earlier in a table and splitting up timestamps and storing their date and daytime components separately. Thanks, Pete [1]http://www.monetdb.org/Documentation/Manuals/SQLreference/Indices [2]http://old.nabble.com/column-sortedness-to33053179.html#a33053179 [3]http://old.nabble.com/General-Questions-on-Columnar-DB-Design-and-Querying-t...
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512 Op 16-02-12 22:38, Pete Black schreef:
Has anyone had any experience with large time series where the column order and types have effected performance? For example having data with less variance and "uniqueness" earlier in a table and splitting up timestamps and storing their date and daytime components separately.
I have seen that the storage of seperate columns only does-good if you in general select on this attribute. So for example searching for a boolean value weekdaymonday goes way better then a function to search for that. Or searching on a 'day' as part of a date. It seems to work against you if you commonly need to compare the contents of all coloms, and if this column fits (normally) in a single register. And my worst enemy seems to be data that is updated ;) Stefan -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.18 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEAREKAAYFAk89hUsACgkQYH1+F2Rqwn2NMgCfXYWb4/nZv7zwAhqlH4MEgWqw 4GEAn1KlarOZX/3q671AjMcRtQl2MDd7 =Ue6D -----END PGP SIGNATURE-----
participants (2)
-
Pete Black -
Stefan de Konink