View Full Version : Generating Aggregations
10-20-2003, 07:34 PM
Like Analysis Manager, how can I generate
aggregations and process them in Mondrian.
10-21-2003, 09:55 AM
Mondrian does not generate aggregations on disk: it builds aggregations in memory only. If you are using Oracle, you can achieve a similar effect using materialized views based upon GROUP BY queries: the results of aggregate queries will be pre-calculated and stored in the database, so they should run really quickly. SQL Server and DB2 support similar constructs.
10-22-2003, 07:14 PM
Ok, but its difficult to manualy make so much
materialized views, if i have 7 dimensions with
average of 5 levels per dimension, then to manualy
make evey materialized view is very laborous.
Will mondrian in future will have such aggregation
10-23-2003, 04:51 AM
I'd like to do some support for stored aggregations, but the problem is what support to build. Bear in mind the following points:
(a) A solution based upon Oracle materialized views would only help a minority of Mondrian users. DB2 and SQL Server have features analogous to materialized views, but MySQL and PostgreSQL do not.
(b) Mondrian needs to work 'out of the box'. It needs to work even if aggregations are not present, and aggregations must update automatically.
(c) Development resources, especially for user interface, are scarce, so this must be easy to implement.
If you have any ideas which meet those criteria, please log an RFE. If you are able to lend a hand to implement these features, even better.
11-01-2003, 04:26 AM
If I manually generate aggregate tables, how do I make Mondrian aware of them?
11-01-2003, 06:32 AM
You can't make mondrian aware of them, and in fact, you don't need to. Mondrian will generate group-by queries based upon the base table + joins to the dimension tables. If you have created the right materialized view, the RDBMS' query optimizer will use that view rather than going to the base tables.