Tikal's recent experience with open source business intelligence, data analysis and report generation tools such as Pentaho CE and many others had proven that the tools are efficient, simple, flexible, scalable, reliable and as free as Linux or Apache.
For some quite time, the Pentaho community (specifically Tom Barber and Paul Stöllberger) is talking about a new web UI for data analysis (OLAP slice-n-dice) and after a few month of hard work they have released version 2.0 of Saiku and it is fully integrated with Pentaho.
Large scale analysis solution must be able to provide very fast response time (<1sec) to some complex queries which are typically structured from a large fact table which is joined to many dimension tables and grouped-by them.
Using standard SQL databases such as MySQL InnoDB (as well as leading commercial vendors) will result in poor performance even when correct indexes are set up since in many case the database will be forced to perform unnecessary full table scans which will kill the performance if dimension tables are large.
Almost all mature database applications reaches a point when it needs to start reporting back statistical and aggregated data. The data about past behavior is usually kept in a transaction history log table and you will be tempted to write some fancy SQL that selects the records in a specific timeslot and then GROUPBY type, or country, or whatever while COUNTing, SUMming and AVGing on some values like profit or whatever make sense to your company.
