Mondrian Logging and Tracing

I have been trying to write up this post for quite sometime now. But lately things have been so busy that I could not find time. So here it is for a lot of folks who have faced trouble setting up Logging and Tracing with Mondrian. Below are the steps to configure Logging and Tracing in Mondrian.

Prerequisite
The below are not the bare minimum but is a tested configuration on my server.

1. Apache Tomcat >= 6
2. Mondiran >= 3.2

Configuring Environmental Variables
While setting up Mondrian we understand that we have to setup CATALINA_OPTS as an environmental variable. For successfully setting up Logging and Tracing, we need to have 2 more declaratives added to the end as below:-



Adding the above in .bashrc in CATALINA_OPTS env variable allows you to do the following:-

• Adding log4j.xml file allows tomcat to pick up the logging configuration from your webapps/mondrian/WEB-INF/classes/log4j.xml. (Sample file is present in the section below)
• Adding log.dir allows you to specify where you want your log files to be generated.

Mondrian, Apache Tomcat and MySQL – Installation and Configuration on Linux

In this post, I am going to walk you through a manual installation and configuration of Mondrian with Apache Tomcat 6. I have always believed to have a clean install from source and thus I am going to follow the same here as well. The following procedure is what I have used in our production and is thoroughly tested in Ubuntu, CentOS and RHEL. Please post your comments/suggestions if you find any flaws in this.

Step – 1: Installing Sun-JDK
There have been many a times that I have faced unknown and hard to debug problems with the default installations of Sun-JDK which comes with Ubuntu/RHEL and other Operating Systems. Specially in production systems, I feel much more comfortable compiling things myself and thus if you are planning to make deployments in production system, it is better that you know what you are installing and where.

Findings on using FlexMonster with Mondrian

Occasionally, you tend to stumble upon a tool which impresses you in the first glance itself. But mostly, these first impressions do not go that far. Recently, for a reporting module, I was looking at a lot of RIA libraries and came across FlexMonster.

FlexMonster is essentially a data visualization solution for any reporting and BI. It's main advantage is having out of the box support for Mondrian(ROLAP) and a Pivot component. Though Mondrian is essentially open source and comes without any cost, FlexMonster solutions costs you about $999 licence. This according to me is a very high price one would pay as compared to the open source alternatives available but some more work into this tool can justify the cost.

One can easily get jazzed by the online demo for FlexMonster but I seriously feel that they have couple of design flaws they need to think through. No doubt, it has a lot of potential to become one of the best OLAP viewers till date, but I am unsure about their development cycle and planned releases if any.

Pros:-

Have tried to lay down some of the things that were strikingly amazing in FlexMonster and made me to like it a lot.

1. Out of the box support for XMLA - This is a serious advantage which many of the other OLAP viewers do not have like JPivot.
2. Exposed as JS library - This makes it really really simple to be embedded into any web application (Python, PhP etc.).
3. UI design is very intuitive - Keeping into mind that most of the OLAP users love simplicity of the UI, it does a fairly good job.
4. Since based on flex, there are no cross browser issues.
5. General Pivot components like sorting/filtering etc. are all out of box.

Cons:-

Here comes the ugly picture that makes FlexMonster unusable for a lot of potential customers.

1. Flexmonster is essentially a Flex version of JPivot but does not do a complete job at it. It wins the race in Rich UI, but lacks a lot of features that JPivot has to offer out of the box for "Free".

2. Lacks MDX support - A lot of business users would want many features that can be taken out only via MDX. Though they claim on the main page of their website - "XMLA and MDX support for multidimensional data visualization".

3. Customization using parameters looks incomplete. For example:- setting the parameter to expand-all as on, it would expand by default on both rows and columns. I would certainly want either my rows or columns to be fully expanded and not the other one.

4. Drill down in charts is only available for CSV and not for OLAP.

5. Stores reports configurations in XML files. This makes embedding the flexmonster as a solution highly unusable in Large Scale data intensive business environments because clients might log in and try to save out their reports. For many businesses, the clients can vary from 100 to millions of customers. This essentially means million or more config files in worst case. Sounds like a bad design to me.

6. Since FlexMonster is reading the config from XML and creating MDX on the fly, their constructs of Rows and Columns should be much more extensive for Developers to user MDX constructs like Hierarchize and Lag etc.

Summing up, FlexMonster gave me a good feel at the first instance, but going ahead its limitations started overgrowing its benefits specially for a production data intensive system. I would love to use this tool again when it matures a bit.