I always wondered what NCR was doing owning Teradata's data warehouse business -- after all NCR makes cash registers, point of sales machines, and all manner of self-service kiosks and the like. Meanwhile, Teradata makes massive sanitary landfills of data, the dumping grounds for terabytes of formerly useful data now rendered largely useless by the sheer size of their new combined mass. Of course, not all data warehouses are bad, just like not all wars are bad. It's just that the problems that both wars and data warehouses set out to solve can all too often be dealt with in a much more simple and cost-effective manner.
So now NCR wants out of the data warehouse business, and so far its shareholders are applauding. If only they knew what a good move that really is. There are three good reasons why data warehouses of the kind that Teradata specialized in are doomed to the ash-heap of history, and NCR is showing uncommon wisdom in acting before the writing on the wall becomes too hard to ignore.
The first problem with data warehouses is that open source is coming right at this piece of the market with both barrels. The most telling example is the Sun/Greenplum data warehouse appliance announced last summer. The price/performance of this little machine has Teradata-killer written all over it. While making the sanitary landfill cheaper won't necessarily fix the underlying problem of building over-large and largely useless data warehouses, at least it won't cost anyone an arm and a leg (more like a hand or a couple of fingers, at worst.)
Another sign of the death-knell of the data warehouse is the advent of in-memory databases and their use in data warehouse environments. Here SAP is the leader, with its in-memory data warehouse offering, also known as the BI Accelerator, or BIA. SAP is doing its own version of a Teradata-killer -- which handily doubles as an Oracle-killer -- by putting SAP's data warehouse on a RAM-based system that vastly outpaces anything the old-fashioned disk and database world of data warehousing could ever accomplish. Again, much more bang, much less buck. See you later, Teradater.
Finally, there's the third way, data archiving. One of the smarter ways to save a bundle on the Teradata model is to off-load your non-essential data warehouse data (i.e. most of it, in most cases) into an archival system that, if well-designed, can be orders of magnitude cheaper and easier to manage as well. One good example of this is SAND Technology's data-archiving solution. Again, the watchwords are faster, better, cheaper. And, in this case, SAND offers a reasonable solution to the sanitary landfill problem: dump the data in a cheap and highly acessible archive, and at a minimum stop paying the big bucks to store information you may never need to user.
Against the backdrop of these growing competitive threats, NCR was smart to start the spin-off, before it's too late. I will be surprised if, a year after the deal is finalized, Teradata's stock will look half as good as it would look now had the spin-off been able to take place immediately. There's a big revolution afoot in the data warehouse market, and the mob is about to put Teradata's head on a pike in the town square. Hopefully for NCR, they'll be long gone by then. Or else.