In George Orwell’s nightmare vision of a wholly totalitarian world, 1984, he highlights the control these dark societies exercise over their populations by the fact that a complete reversal of policy – we are at war with Eurasia, and Eastasia is our friend, then we are suddenly at peace with Eurasia, and Eastasia is our […]
In George Orwell’s nightmare vision of a wholly totalitarian world, 1984, he highlights the control these dark societies exercise over their populations by the fact that a complete reversal of policy – we are at war with Eurasia, and Eastasia is our friend, then we are suddenly at peace with Eurasia, and Eastasia is our enemy – gets announced without a murmur of dissent. One might be tempted to think a similar 180 degree reversal of policy is behind Menlo Park based data mart specialist Informatica Corp’s announcement that it is all in favour of data warehouses. Given that the company, best known for its Unix and NT products called Powermart and Powercapture, loudly proclaims it was the mainspring behind the 18 month old trend for companies to avoid the Big Bang or ‘galactic’ data warehouse – where a company attempts to build a enterprise wide data model and repository first, off which it can download data marts as necessary – and instead build smaller-scale subject- or department-oriented data marts, this on the face of it does seem surprising.
By Gary Flood
For the data mart phenomenon has introduced much controversy between data warehouse purists, often quoting self-styled father of the data warehouse Bill Inmon, and the data martists. Inmonites allege that data marts encourage the old IT problem of ‘islands of automation’ – whereby data marts may get set up across an organization, do useful stuff and have more data added to them, but because of different definitions of core data assumptions – the metadata problem – become too motley to intercommunicate. Data mart folks counter that the Big Bang approach ends up as far too costly and long-winded a method of delivering the intended business benefits behind a souped-up decision support database, which is what data warehouses/marts are at the end of the day, and that data marts are the more pragmatic solution. Now Informatica has announced the Enterprise Data Mart architecture at a data warehousing trade show in Boston this week, at the moment little more than a vision statement, but upon whose foundations it intends to release its next generation of products and services, code named Tornado, later in the year. According to Diaz Nesamoney, president, chief technology officer, and co-founder, 30% of its 90 customers are in production mode in just six months, with some sites hitting that state in between only 60 to 90 days, and, in a somewhat embarrassing bit of hyperbole, euphoria abounds as a result. But these successful customers, says Informatica, still face issues – data marts are as discussed still largely being seen as data islands, achieving satisfactory data quality is a constant battle, there is the aforementioned need to integrate data models for consistency, and there is the nagging worry that there is all too often no overall architecture guiding the architects of the data marts.
Right architecture, wrong implementation
Thus the company admits that in many ways the data warehousers have the right architecture, it’s just that they have the wrong implementation. Informatica claims the happy medium with this architecture, which would seem to reconcile both Inmonites and data mart fans and offer a means to solve the integrity/consistency versus affordability/length of implementation dilemma. By claiming that the EDM (enterprise data mart) unites the top down and bottom up approaches, we have at last a practical incremental approach to data warehousing. Well, splendid. But it must be pointed out that Informatica is not as yet offering anything new, and this is just a placeholder statement of direction only. All we know at the moment is that the EDM will offer new software in terms of what is currently being called the Dynamic Data Store (which Nesamoney says is conceptually new, though it’s actually just physically a customised standard relational database application), and the Global Data Mart Repository, which will offer a complete metadata catalogue of the dynamic data store, allowing metatdata sharing among the data marts, plus an extended management framework. And it must also be pointed out that there are many other data cleansing and scrubbing vendors, such as long time rivals Prism Solutions Inc, which also offer their own competing grand visionary architectures, albeit it normally at the higher end. Informatica believes the most receptive market for the idea will be consultants (it has an OEM agreement with KPMG), systems integrators and its own expanding customer base, arguing that this move will extend its leadership position (at the more cost- conscious end of the data mart spectrum, specifically). Informatica alleges it is making great hay with its data mart approach than its galactic rivals, with Paul Albright, vice president of marketing, sniping that we have less shelfware and more customers in production than rival vendors. The move coincides with the firm starting to ramp up for its IPO. The company received an undisclosed amount of start up capital with an October 1995 round of VC (venture capital) money, from Bay Partners, Discovery Ventures and Partech International, the VC arm of France’s Paribas Group. Only last month (CI No 3,212) it raised $9m in mezzanine (last round prior to going public) from those investors and a new VC partner, Weiss, Peck & Greer. Is there not a touch of Eastasia-Eurasia in all of this? Isn’t this a gift for Informatica rivals, who can now claim that even Informatica has come around to the data warehouse top-down idea? Not true, says Nesamoney. We still believe it’s wrong to try and do consolidation and customization at the same time, and that you need to break it into different tiers – and this approach is still incremental, and more rapid than others. In any case, it will be a hard sell for data mart vendors without any kind of grand vision framework now that even Informatica has acknowledged the importance of such an architecture.