For my next DWBI project, I’m thinking about implementing Oracle 11g OLAP option instead of SSAS, so I’m comparing them.
- Browsability: users can use OBI (e.g. Answers), BO (since 9i AW/2003), Excel or other tool to browse Oracle OLAP. Users are not expected to use SQL to query the cube. In AS now we have Crescent to look forward to. In the mean time we could use Tableau (see Jen’s session in SQLBits 8), Strategy Companion, CubePlayer or Excel 2010.
- Oracle OLAP cubes are stored as true multidimensional store (I’m interpretting this as compressed MOLAP), not as relational tables with materialized view/MV. This MOLAP storage solves the classic ROLAP problem of “MV aggregates can’t cover every permutation of every member of every attribute against other members of other attributes”.
- Oracle mentions compression & cost-based aggregation (see page 7). SSAS is certainly compressed, but I’m not sure if SSAS aggregations are cost-based.
- Oracle OLAP has the ability to blend multidimensional and SQL models within the same query (see Bud’s comments below). This “blend concept” is similar to SQL Server’s BISM concept in Denali (2011). BISM can expose itself as multidimensional interface (MDX) or SQL/relational interface (DAX). The difference is that BISM is relational (tables), with “in-memory column” (VertiPaq), where as 11g’s OLAP is compressed MOLAP.
- Many-to-many relationship. If you are not familiar with M2M in SSAS, pls read Marco’s paper about M2M Revolution. This is critical for many business situations. Can’t live without it. Have searched the whole library and I couldn’t find M2M in 11g OLAP. Bud said we can do it (see below), and from what I understand here, this is done by simply mapping a base measure to the fact table containing the M2M relationship, just like the way we create any base measure. Not sure how do we hide the measure though (because the measure is dummy, it is only used to establish M2M relationship between 2 dimensions).
- Yes 11g OLAP Option can create calculated measures. But as we in SSAS world know, calculated measure such as YTD or growth is not a problem if it’s only 1 or 2 measures, but if we have 50 measures we use date tool aka date calculation dimension (see Chris, Alberto & Marco’s most useful SSAS book here, chapter 6). Seems that we can’t do this in 11g OLAP. Surely we can’t create 50 * 10 calc measures? Well we can, but how about the usability?
Update 27/12: I have read Bud’s comment below, and I know we can create “measure dim” and “calculation dim”, but in SSAS we need to create a hidden measure and write some “MDX calculations” to establish how the growth/2 yr growth, etc will be calculated. Have read about creating calc measure on 11g here, but couldn’t find how that “MDX calculation” is done in Oracle OLAP. Could anybody point me a direction?
- Complex calculation/MDX, i.e. existing (results depends on what’s on the slicer). Hidden named sets which is used by a calc member, etc. How do we do that in 11g OLAP?
Update 27/11: have explored Simba’s MDX Provider for Oracle OLAP (link, link, link) and could not find comprehensive OLAP/MDX functions such as this. It seems that Simba’s product only provide connectivity, not ablity to do complex calculations. Would be grateful if somebody can enlighten me (list of Simba’s MDX functions for Oracle 11g OLAP). Have also looked at Bizzantz’s DeltaMaster methods, but it’s far below the richness of AS’ MDX e.g. scoping, existing, etc.
- MDX interface. 11g OLAP cube does mention MDX interface to OBIEE, but I expect this is basic “select … on columns, … on rows from cube” without the ability to use MS’ MDX functions such as NECJ, parallel period, etc. See page 14.
Update 27/11: same comment as point 7. Update 28/11: see OLAP DML Functions here.
- Measure expression. This is required for example for currency conversion. If we don’t have measure expression in 11g OLAP (and we don’t have M2M either), how do we do currency conversion?
Update 27/11: we can do this in 11g OLAP using “calculated measure” (see here), just need to specify the correct “calculation type” and “calculation”. It is quite comprehensive functions with regards to period / time series computation, but outside period I’m not sure. Update 28/11: found it. It’s called OLAP DML Functions, see here. It’s very comprehensive.
- Dimensional/cell security. In AS I sort of take it for granted that group1 can only see member1 whereas group2 can only see member2, along with giving users different default members based on role. But in 11g OLAP?
Update 28/11: as Bud said, yes we can do dimensional security in Oracle OLAP, see “Selecting data by criteria” here.
- In AS we can control the incremental cube processing, i.e. process changed partition only etc. It is mentioned in 11g OLAP this can be automated (see the end on page 28 re Fast cube refresh), and that they can update incrementally by reading the log files (see last paragraph of page 6). This is a good thing (compared to AS).
If you find something incorrect above, I’d be grateful if you could let me know, either via comments or firstname.lastname@example.org. I will correct them. Thanks again for Bud Eggress for your corrections.