In response to the following help request from a Cambridge Technology Partners alumnus and friend...
>> I'm looking to flatten a vertical structure into a horizontal one for reporting purposes...
I replied,
>Let's talk. The old model of relational database reporting is dead.
>There is a better, faster, and cheaper way using an unstructured data model cloud analytics/reporting solution.
This sparked a flurry of responses including:
(1) Weather report for Cambridge Massachusetts today: Cloudy with a chance of unstructured data.
(2) The relational database is dead...God save the relational database!!!
(3) Having lead several Datawarehouse projects, I can honestly tell you that my team easily spent 30-40% of the versioning RDBMS schemas to model the typical transformation pipeline (star schema). When we switched the same process to Hadoop were able to go against the detailed records, w/o pressing them into the relational model and just reprocessed them due to the massive amounts of computing power we had. Our $500k Oracle/EMC datawarehouse was replaced by $50k worth of Hadoop cluster machines. Runtimes went down from 20 hrs. to 30min for our particular case (telco mobile metrics). I'm definitely a convert and not going to touch RDBMS' for Datawarehousing anymore.
At Cloud Technology Partners, we help transform client's reporting infrastructure with two great solutions.
1. The Hadoop model combined with a reporting solution when
- report data must remain in the client's own data center
- the client can use pure unstructured data
2. A secure cloud-based business intelligence solution which
- pulls in unstructured data from systems
- establishes flexible meta-data relationships in the cloud
- provides a slick web front-end for rapid report creation
- has a low TCO (total cost of ownership)
For more, see our review of PivotLink.