Hi,
On 12 October 2011 14:50, Anthony Presley <anthony@resolution.com> wrote:
> After a few weeks of searching around, we're running into dead-ends on the
> front-end, and the back-end. PG doesn't support OLAP / MDX .... and the GUI
> tools that do this, for the most part, require MDX / OLAP (SPSS and DB2, MS
> SQL Analytics, etc...).
> What's the PG route here? Are there some secrets / tips / tricks / contrib
> modules for handling this?
Our db has very simple star schema and several materialised tables for
aggregates. We need OLAP style queries with OLTP performance :). After
several upgrades we ended up with 16 cores, 80GB of RAM and fast SAN
but performance wasn't good. Meanwhile we evaluated several NoSQL
solutions and several comercial MPP solutions -- and the winner is
Greenplum! (you can get CE version for free -- max two processors (but
unlimited cores) and almost no HA options). AsterData's nCluster has
very nice MapReduce integration but looks quite immature.
I would say that roll-ups/drill-downs/cubes sound obsolete to me :) I
suggest the following tools (list is not sorted by anything)
- RapidMiner (http://rapid-i.com/content/view/181/196/)
- Tableau (http://www.tableausoftware.com/public/gallery)
- AlpineMiner (http://www.alpinedatalabs.com/, http://alpineminer.org/)
- Jasper (http://www.jaspersoft.com/)
- Pentaho (http://www.pentaho.com/)
--
Ondrej Ivanic
(ondrej.ivanic@gmail.com)