mondrian-olap 0.2.0 is released

[Raimonds Simanovskis]

Just released mondrian-olap version 0.2.0 with several bug fixes and improvements (that were done since the initial release) as well as with support for LucidDB database. See changelog at https://github.com/rsim/mondrian-olap/blob/master/Changelog.md

In last days I am experimenting with LucidDB (http://luciddb.org) as backend for Mondrian OLAP and when I have many millions of rows in fact table then I start to see performance improvements over traditional row-based databases. So I recommend to check it out if you have large data warehouse databases with performance issues when doing analytical SQL queries.

Raimonds

[Luc Boudreau]

Raimonds,

The general rule about backends for Mondrian is that "analytical" databases always outperform their "transactional" counterpart. RDBMS like Oracle are slower at analytical type queries than LucidDB or Greenplum. Also, we recommend staying away from MySQL and other "commodity" databases as their performance is usually outright unacceptable for analysis. We have also had amazing performances using Vectorwise or Infobright.

Just my 2 cents.. cheers!

Luc

[John Gordon]

Apologies for this being off-topic, but I need to ask a question that I'm assuming everyone knows already: How is a jruby rails app deployed in prod? I ask after trying a few approaches that do not quite satisfy:

Phusion Passenger -- Wow, does that not work. I don't see the documentation saying that Passenger doesn't support jruby rails apps, but I've come to the conclusion from my own experimentation.

Warbler -- kinda, but it means giving up on Capistrano

Trinidad -- looked really good, but i'm seeing major performance issues. a typical query is just choking the app.

What's the best way?

My Prod environment is CentOS.

Thanks, and sorry for the off topic nature of this one, but I'm stuck.

John

[Raimonds Simanovskis]

I am currently using Kirk (https://github.com/strobecorp/kirk) which is more lightweight than Trinidad.

In front of it I use Apache which serves static assets and then proxies other requests to Kirk in backend (similar to e.g. Unicorn deployment when using MRI).

Raimonds