A brief look at Trillium's Discovery 5.0

I recently received an update from Trillium Software on its latest release (5.0) of Discovery, the data profiling and analysis product it acquired from Avellino.

Even though the product is all about profiling and analysis there are actually no new profiling and analysis capabilities (apart from ease of use and similar things – including selective profiling, which allows you to profile only what you want) in the product.

Why is this? Simply that there is only so much profiling that you can do, and Discovery does it all. This, of course, raises the point that sooner or later (probably sooner) other products will be in the same position. So how can Trillium keep ahead of the trailing pack?

Trillium started to answer this question in its previous release with support for user-defined business rules and it has extended this capability further in its latest version, so that you can now apply business rules to do things such as create calculated columns or concatenate fields.

Now, if you think about it, these are transformations in the same way that ETL (extract, transform and load) applies transformations. And since Discovery can read data from source systems and output it again to a target this means that Discovery could actually be used for ETL purposes.

Of course, there is no job stream definition or scheduling (though the latter could be implemented externally via the command line capability that has been introduced in this release), and Discovery is more aimed at business users than the developers that do the bulk of the work with conventional ETL tools but, for simple jobs, it is certainly capable of being used in that way. Note, however, that this is not a target market for Trillium: it would not want to upset its partners in the data integration space.

Business rules have also been upgraded with the ability to apply them, dynamically, to external data. That is, you can apply business rules without having to pull the (external)data into the Discovery Metabase (see below). In fact this, in a rather different way highlights the general thrust of release 5.0, which is very much focused on the external environment. Also, there are enhanced abilities to embed profiling in third party environments, extended integration with Trillium’s data cleansing and matching software, and an email notification capability, amongst others.

Finally, and this is not part of the new release per se, Trillium told me that it still gets grief over its metabase (where it stores all its data and metadata), because it does not run on top of SQL Server or Oracle or some other popular relational database. Let me state my opinion here firmly: this is a very good thing.

The point about the information in the Discovery metabase is that it does not change: you may enter new data and you may purge old data but there is very little, if any, updating of data. In other words a database designed for transactional processing is not appropriate. In fact, Trillium uses BerkeleyDB from Sleepycat Software for its metabase, which is entirely appropriate—it should provide better performance with less overhead and reduced administration—which is as good a set of arguments as it gets for not using a relational database.

Copyright © 2006, IT-Analysis.com