As 1997 draws to an end, it does seem to be time to start to wrap up 1997 in some ways – we can begin to look back on some of the debates that have exercised us all this year with a little more perspective than when it began. Take the Data Mart versus Data Warehouse controversy. If you buy the Data Warehouse idea – that data, sucked out of its operational data rind, should be collated and suitably stimulated so as to provide a decision support database to explore what-if scenarios and the like – then there seems to be two ways to get there. Either you go hell for leather for building a Data Mart, i.e., a subject-specific or one-line-of-business-only Data Warehouse, or you start planning a more overarching corporate-wide Warehouse. The criterion for deciding for a Mart usually revolves around the need to get something up and running, and (naturally!) delivering business benefit in a short a time as possible. The basis for opting for the more centralized Warehouse depends usually on the wish to avoid inconsistency and duplication, as Mart begets Mart and they begin to struggle to talk to each other. Put like this, the choice seems one predicated purely by the amount of available time – plainly, if one had the time, one would build a warehouse.
By Gary Flood
What often gets missed in the debate at this level are two factors that must also be included. The first is the nature of the organization seeking to build the system in the first place. It doesn’t take much reflection to agree that if it is an enterprise with a central IT resource, which also controls the budget for any such system, it will tend more naturally by weight of intellectual rigor and caution swing to the Warehouse. These people, sweating over the Year 2000 and the European Monetary Union thunderbolts, know too well the price their companies – and ultimately they themselves as the grunts most likely to have to clean out the corrupted data foxholes – will pay for launching a business critical system that runs the danger of becoming rapidly too difficult to maintain, and eventually, trust. (Remember expert systems? One of the most important cultural obstacles to their acceptance was that too few of us were willing to believe the computer was that smart, since we were not privy to its train of thought, and hence didn’t buy off on the eventual answers.)
Modern business sociology
By the same token, modern business sociology, post-BPR, gives us a typical picture of a company with lots of internal, semi-autonomous business units, perhaps as competitive with one another internally as they are externally. These individual business unit managers will gravitate to solutions, as opposed to home brew, since this is simpler and quicker to implement. So there is the cultural dimension to the Mart versus Warehouse debate. There is also the fallacy that a Mart is axiomatically a small thing, an illusion fostered by the fact that conceptually it is a subdivision of a larger thing, the Warehouse. For the strict definition of a Mart is simply that it is a subject-oriented Warehouse – not that it is a small database. There is absolutely no reason why one couldn’t have a 1Tb Data Mart, if all it did was hold information forming a constellation of views around one particular problem. What is certain is that users seem to like both Marts and the idea that, if it were all possible, having some responsible party assemble all the different elements – for there is no such thing as a piece of Data Warehouse software, it being composed of many things, and equally there is no one vendor that does all of this stuff. For example, according to a Robertson Stephens & Co report last year, as much as 50% to 75% of the whole $15bn decision support software market will be Mart-based over the next five years, with Sentry in 1996 finding 53% of those surveyed would prefer a single vendor solution. These trends form an important part of the background to both Informatica Corp’s recent attempt to place a stake in the ground, whereby the Informatica Data Mart can successfully grow up into the Informatica Data Warehouse (CI No 3,283), but is also behind Oracle Corp’s move into the Data Mart area with a packaged suite for Sales & Marketing. Oracle entered the Data Warehouse space just under two and a half years ago, in June 1995 (CI No 2,695), and dipped into the Mart area at the end of March this year (CI No 3,131). The company finally got around to releasing a generic version of the Data Mart suite in September. Now the Sales & Marketing version is due later this month. The idea is that, on one CD-ROM, a customer gets all he needs to build a Data Mart, including design and front end tools, and the database management system itself. Oracle has decided not to lower the price of the offering by slotting a scaled down database. Most of the technology in there is from Oracle, it claims, though it adds that in order to offer a complete solution the company licensed Data Flow extraction technology from Palo Alto data movement specialist Sagent Technology Inc. Oracle claims it is leading the market with the such packages, arguing that instead of bundles, a customer gets one box, with one Oracle 800 number to call if he has a problem, and a cookbook [methodology] to get it up and running quickly. The Redwood Shores database giant also claims to be the first out of the starting gate with a Mart suite, as opposed to a bundle. (It may be, but merely by a nose – Sagent itself announced in September an HR suite built with computing services giant Automatic Data Processing, Inc, available sometime before the end of the year, and there are bound to be others in the pipeline from other vendors.)
Cuddly little database
Will MSFT – whose SQL Server on NT is reckoned the prime Data Mart database engine combination – be able to compete by pointing out that the Sales & Marketing suite, at just under $100,000, makes a mockery of the whole idea of cheap and easy to implement Marts? A Data Mart is not some cuddly little database that I can buy and let happen – it requires some industrial strength scalability, as customers inevitably add more and more data to it. Oracle – surprise! – claims to be the Mart leader on NT over SQL Server as well as in the general Warehouse space, and that SQL Server Marts over 50 Gbytes tend to be flaky. There is bound to be a market for this product, and others similar, but in and of itself it won’t solve the Marts versus Warehouse problem. One suspects that as ever with IT, we shall rush in with little gleeful cries of excitement (just think how much legacy Java code we’ll all have soon, written by programmers who wanted an excuse to learn it between 9 to 5), build a bunch of Marts all over the place, and have a big problem to face sometime soon. But that won’t be until 1998 at least, when we should also have other controversies to chew over.