IBM's (and Oracle's) plans for U2 users
IBM better get its skates on
Briefing Note I recently had a briefing with IBM during which the subject of the future of Informix users came up, writes Phil Howard. The answer to that question is well known: IBM will continue to maintain and upgrade the Informix database for the foreseeable future but, in the longer term, IBM expects to provide migration capability to DB2.
Mostly as an aside I also asked what plans the company had for its U2 databases, which it also acquired when it bought Informix. The answer was that it had no specific plans as the U2 databases use fundamentally different technologies from DB2. I thought this answer a little odd but did not pursue the matter.
I had better explain what the U2 databases are and how they are different. The U2 databases consist of UniVerse and Unidata (hence U2), which were originally developed by VMark and Unidata respectively. Subsequently, the two companies merged to become Ardent Software and this in turn was bought by Informix, primarily for the company's DataStage ETL (extract, transform and load) tool.
UniVerse and Unidata are relational databases with a difference. In conventional relational environments the first rule of relational technology is that you cannot have something called a repeating group. This would occur, for example, if you stored order header details and then all the order line details that pertained to that order within the same database table. In a relational database the order header details and order lines are stored in separate tables. Technically, this is known as First Normal Form. All standard relational databases adhere to this principle.
However, as it happens, relational algebra (which is based on set theory) does not actually require adherence to First Normal Form. If you are careful enough about your implementation then you can preserve the integrity of relational algebra while allowing the use of repeating groups. This is what the U2 databases do and, for this reason, they are sometimes referred to as NF2 databases (Non First Normal Form).
NF2 has two big advantages. The first is that you get better performance. By storing order lines with header details you can retrieve all relevant details from a single table rather than multiple tables. Secondly, you have far fewer tables. This makes the organisation of the database very much easier to understand and manage. Not surprisingly, therefore, the NF2 databases have proved very popular with VARs and software houses that want an easily managed, "fire-and-forget" database that they can embed within their software packages. As a result, the U2 databases are leading providers within the embedded database market, behind Oracle and Progress.
What surprised me somewhat about IBM's response to my question about the U2 databases was that, actually, these are not so far apart from DB2 as you might imagine. This is because DB2 supports something called structured types. Without going into detail about these are, it is only necessary to know that you can embed structured types inside other structured types.
In other words, DB2 supports nested tables, which is an exact equivalent of the NF2 approach used in the U2 databases: you construct a structured type for your order header and embed a structured type for order details within the first structured type and voilà. Obviously it's not quite as simple as this but the possibilities for migration are certainly there.
To be honest, I would not have bothered in commenting on this lack of vision had I not had a recent briefing with Oracle, during which we discussed that company's migration strategy for Informix users. Almost in passing the subject of the U2 databases came up. And lo and behold, Oracle has a major U2 software house migrating all of its applications to Oracle. And how are they doing that? By using the nested table capability within the Oracle database.
According to Oracle, the company has been pleasantly surprised at how simple and quick the conversion has been. You might argue that Oracle would say that wouldn't they, but actually it is logical and not particularly surprising.
The bottom line is that IBM had better wake up or it is in danger of losing a substantial number of its U2 customers to Oracle. Once word gets out that this migration is relatively easy, significant numbers of U2 software houses may make the switch to Oracle.
Phil Howard's Biog
Sponsored: Hyper-scale data management