InformationWeek
Defining The Business Value Of Technology
| September 15, 1997 | |
Dependent Data
Enterprise data marts linked to data warehouses may make independent data marts obsolete
By Teri Robinson
Informatica, NCR, and Oracle have each launched dependent data mart strategies. Last month, Informatica offered a blueprint for constructing and networking data marts into enterprisewide data warehouses with its enterprise data mart architecture. This approach helps prevent data islands, says Paul Allbright, VP of marketing at the Menlo Park, Calif., company.
The enterprise data mart is where many users are headed, notes
a June report from the Patricia Seybold Group, a consulting firm in Boston. Seybold puts data mart solutions into two categories: Data Mart Lite and Data Mart Enterprise. The lite category covers solutions that incorporate multiple data warehouse components into one product. Microsoft, Broadbase, and Sagent Technology offer such products, the report says. The enterprise category applies to solutions that have features to enhance the overall functionality and performance for high-end installations. Enterprise vendors include the likes of Prism Solutions, Carleton, Informatica, Information Builders, Vmark Software, SAS Institute, D2K, and Evolutionary Technologies International.
Informatica's and other enterprise data mart solutions offer the quick return on investment of independent data marts as well as a growth path to the future. Users can build data marts incrementally, then repeat the success throughout the enterprise, says Allbright.
What's more, the architecture offers the best of both the cent
ralized and distributed approaches to data warehouses-IT can centrally manage the marts with a consistent corporatewide view of information, but the business logic resides in the data mart. "Companies aren't anarchies, so why should their data models be anarchies?" asks Allbright.
At Owens & Minor, a $3 billion distributor of medical surgical supplies in Alexandria, Va., the initial project is being thought of as an enterprise data mart, says Don Stoller, director for decision services. The company needed a mart to distribute sales information. Owens & Minor's data mart runs Oracle7.3 on a Hewlett-Packard T520 server and uses business objects for end-user query and Informatica's Power-Mart for extraction and loading. It runs in 40 different distribution centers across the country. The company has realized a 15% sales increase to one group of hospitals-and that should boost 1997 revenue by $44 million.
While some users have had great success over the past year with data mart implementations, others
have learned hard lessons about scalability-or the lack thereof. They sank precious data warehousing dollars into quick solutions wrapped in big promises that can't keep up as business flourishes and their information needs explode.
"In a lot of these things, what you get is some basic reporting tools, but is there really any business value?" asks Stephen Cranford, a partner at KPMG Peat Marwick in Radnor, Pa. "There has been a lot of disappointment over these project implementations."
For many companies, the quickest returns come from the so-called data marts in a box, prepackaged solutions that offer all the elements-including tools needed to construct a mart. These are sold by vendors such as Informix, Sybase, and Pilot. Their promise: A data mart that's up and running in 60, 90, or 120 days. That's as close to an instant return as business can get, says Cranford: "The ROI is very quick."
But is it as easy as all that? If a company needs only a basic data mart-a reporting function, for example-then a data mart in a box is probably safe. "It's worth your while to do it quick and dirty if you're just connecting to a reporting database," says Patrick Nolan, an MIS engineer at Stanford University in Palo Alto, Calif.
For Stanford and other users that requ
ire a more customized solution, prepackaged solutions won't do. The Canadian government's Net Income Stabilization Account (NISA) Administration, part of Agriculture and Agri-Food Canada, briefly con- sidered a data mart in a box before deciding to go with a best-of-breed approach, says Mark MacIver, a technical architect at NISA, in Winnipeg, Manitoba.
As the lure of the packaged data mart has grown, so have the ranks of vendors. Nearly every vendor-from generalists such as Microsoft to leaders in the 1-terabyte-or-larger data warehouse crowd such as NCR-has a data mart strategy. Most claim to have canned solutions that range in price from $50,000 to $100,000-significantly less than buying all the pieces separately. But, just like their creators, not all boxed sets are created equal.
Some solutions are physically incapable of growth. "Scalability is very important," says NISA's MacIver, who puts a special emphasis on processing power. "We've got a lot of room to grow on the hardware side as well
as on the software side." NISA has an RS/6000 J40 with four processors and is using Sybase SQL Server 11 and Sybase IQ. PowerMart sits in the middle.
Seeing Is Believing
Alan Paller, director of education at the Data Warehouse Institute in Gaithersburg, Md., notes that doing a test drive isn't enough. "There are a lot of bottlenecks that don't show up when you have five to 10 users," he says.
And not all boxed sets offer the same components or same level of service and support. Some sets provide consulting services only during the 60- to 120-day implementation period. Also, a canned data mart can require a lot more vendor hand-holding than that-unless an IT department ha
s the talent and the time to do it all in-house. "We're stretched too thin to do it ourselves," says one IT director who asked not to be named.
Equally important to scalability is the vendor behind the solution. While the data mart market has attracted a lot of good vendors, says Allbright, it has also drawn in "a lot of charlatans." A few vendors have pieced together sub-par products or simply lack the support system to cater to potential customers. "The problem is that everyone is jumping on the bandwagon to do these things," says Cranford, who notes that data marts can be "tactical cash cows" for vendors and consultants.
Other vendors are clearly focused on the short term. While there's nothing wrong with this strategy, these vendors may not be able to meet the needs of a growing enterprise. "We think the distinctions between Data Mart Lite and Data Mart Enterprise will grow over time," says the Seybold report.
As the data mart market has grown, it has become increasingly clear that some v
endors are slow to make improvements and upgrades to their data mart solutions. The number of upgrades offered-whether they are full or partial upgrades-and the numbers of customers and products a vendor has reveal a lot about how capable a vendor is, says Allbright. "We offer up new technology in cycles fast enough that we're one step ahead of our customers," he says. Informatica recently aimed its PowerMart 3.5 software suite at the growing data mart environment.
Road Map
The Cognos program has already helped York International increase the volume of information that its sales force analyzes, says Gordon Kahn, VP of corporate finance. York uses two different
ways outlined by the program. When it recently acquired a company, York was able to quickly reformat that company's database to mesh with its own.
A growing number of vendors are slowly realizing that they must boost the value of their canned data marts or face disgruntled users. Over the past year, data marts in a box have improved steadily, says the Data Warehouse Institute's Paller, with some vendors adding intelligence to their solutions.
KPMG's Cranford suggests that the vendors who best meet user growth needs are those "who will own the marketplace, vendors like Oracle and Informix." But he also points out that specialty vendors such as Treasury Management Services and PeopleSoft also offer solid-and growable-data marts.
But even with the right vendor and a growable solution, the data mart in a box may not spring to life without a struggle. The time it takes to get a mart going will depend in large part on the condition of the data going into the mart. "In a perfect world, where the data is
clean, a mart can go up quickly," Cranford says. "But if the data is in bad shape, you'll have to spend time cleaning it up."
Big Picture
Without a grand design-a plan to grow a data mart and link it to an overall corporate data warehouse strategy-a company can hardly expect a data mart in a box to grow beyond its initial boundaries. "You can't build them in a vacuum," says Stoller, who is developing a data mart that will integrate over time into an enterprise data warehouse
.
Vendors such as Informatica, NCR, and Oracle are nudging users in that direction with their respective Enterprise Data Marts, Dependent Data Mart, and FastForward offerings. All three are built around the concept that a mart should ultimately depend on the overall data warehouse-in NCR's case, a Teradata implementation. They've also priced these data mart solutions to be attractive to smaller businesses or departments. Pricing for NCR's Dependent Data Mart, for instance, starts as low as $37,000.
Planning Is Key
It took Stanford University two years to plan its data mart, but the school has plenty of room and flexibility to grow. Stanford has "a few IBM 3090s and about 50 Sun machines," says Nolan. But the university is moving to a Sun platform-a Starfire
high-end server has been purchased, but isn't running yet-and will segue from Sybase to Oracle.
Taking small bites may be a better strategy than plowing into a full data warehouse, but one taste often leads to a hunger for more. IT often finds it hard to satisfy that demand with a rudimentary data mart in a box. Still, "data marts are a very good start-even data marts in a box, if you pick the right one," says KPMG's Cranford, who predicts that vendors will only make the solutions more scalable and smarter soon.
|
he independent data mart may soon be a thing of the past. Its replacement: dependent, or enterprise, data marts that "network" into an overall data warehouse strategy.
Featured Microsite
