Are OLAP cubes dead?
Hi everyone,
it’s been a while since my last post, many things have changed since then see it as an another life.
Well today i was thinking: “Is there any chance that OLAP cubes are dead?”
You’re certainly are saying: “What’s wrong with you? I’m using them obviously are not!” and i’m not sure about this.
What are OLAP cubes?
OLAP cubes are the last stage of data analysis. Once you have defined your data warehouse ETL and tables then or you can stop here or you can develop OLAP cubes which are a sort of particular “point-of-view” of your fact.
- A cube has dimensions (Dimensions tables)
- A cube has a fact (Fact table)
- A cube can drill through data
- A cube can slice data
- A cube can be shrinked
Well even simple relational database can do this.
PostgreSQL integrates since version 8 some very useful big data functions: https://www.postgresql.org/docs/devel/static/queries-table-expressions.html#QUERIES-GROUPING-SETS
Oracle performances are tuned for this kind of operation:
http://www.oracle.com/technetwork/issue-archive/2007/07-jul/o47bi-093416.html
Well major databases developers are introducing more function to make life easier to BI analysts and data scientists instead to make compulsory the use of OLAP cube.
Are those cubes really bad?
Well no, they’re simply staring to be a little old, now Amazon web services, Google Cloud Platform and other major SaaS are giving machines with fast SSDs so classic query performs really good on a huge amount of data with a not optimised query, OLAP cubes were developed for sake of simplicity and to avoid lack of performances on queries.
Even Microsoft at first were deleting OLAP cubes support on their Power BI suite :https://community.powerbi.com/t5/Report-Server/Non-analysis-services-data-source-connection-to-power-bi-and/td-p/250088
but then… https://jetsupport.jetreports.com/hc/en-us/articles/218952488-How-to-Connect-Power-BI-to-OLAP-Cubes
They added the OLAP cubes support.
Nowadays servers are really fast and the connections are fast as I/O throughput, many technique were developed such as “in-Memory” or “real-time” which gives immediate results even in really big tables in no time.
In conclusion i think that cubes will die soon or later with the born of new functions that replicates what a cube does without the construction of an another data support, data retrieval is equally fast between this two worlds.
What do you think? Leave a comment below.
