There has been a lot of comment recently (not least in this column) about data warehouse appliances. Unfortunately, much of that comment is either uninformed or biased, or both. In particular, those with a vested interest in the matter have been creating a lot of fear, uncertainty and doubt (FUD) about the whole concept. While I plan to address this in detail in a future white paper, I would like to take this opportunity to discuss one of the most prevalent put-downs: that appliances are proprietary.

Now, the first thing that you might notice on any article that is attacking the idea of appliances is that Teradata (if it is mentioned) always appears as one of the good guys and it is never classed alongside the appliance vendors. This is odd given that Teradata uses a proprietary database, with a proprietary interconnect, with a (near) proprietary operating system in an environment that is more or less wholly and exclusively proprietary. I guess the idea of associating the new kids on the block with a success story might not be what the authors of these articles had in mind.

Of course, and as an aside, the new generation of appliance vendors mostly do not compete with Teradata in the enterprise data warehouse space yet and this could be used as an excuse. But note the use of the word “yet”. They may not now but they will do in the not too distant future. Indeed, there are already appliances being used as systems of record.

Anyway, to return to the point, it is generally conceded that appliance vendors today rely (to a greater or lesser extent) on commodity components (Intel processors, open source databases, standard interconnects and disk arrays and so forth) but the argument goes on that upgrades are not as easy as on conventional systems and that their proprietary nature remains an obstacle because you cannot reuse these components for some other purpose.

As far as upgrades are concerned, I don’t see that, in principle at least, there is anything different between upgrading the hardware in an appliance and upgrading a processor from, say, Sun. Of course, implementations vary and upgrades may be more or less easy to do but that’s a vendor issue rather than a generic one.

The reuse of the components argument seems a futile one to me. Firstly, a typical appliance architecture has an SMP system at the front-end—can this be re-purposed? Certainly, and the same is true of the disks. In addition, depending on the supplier, appliances can be used for different tasks within the warehousing environment and are not necessarily limited to a role as a data mart.

However, much more important is the question as to whether this reuse argument is actually relevant. After all, by the time you might want to re-purpose (if you ever do? when was the last time you did?) how old is the hardware going to be? (And, if it’s not relatively new, what were you doing buying it in the first place?) In any case, wasn’t the software the biggest part of the investment you made?

My bottom line is that if it does what you want it to do, at a price point that you are prepared to pay, and subject to service levels and a forward migration/upgrade path then it doesn’t matter whether the solution is proprietary or not: the whole issue is entirely notional and the latter day equivalent of debating about how many angels can dance on the head of a pin.