 |
Articles
|
||||||||
|
Web browsers can be a cost-effective way to provide customers and employees with access to a data warehouse. By Neil Raden Issue Date: May 13, 1996 Over the past 12 months, the two most pervasive themes in computing have been the Internet and data warehousing. From a marketing perspective, a marriage of these two giant technologies is as natural as peanut butter and jelly and as riveting as Caesar and Cleopatra. Not wanting to miss out on the latest hoopla, vendors are lining up with previews of products-some long before they are ready-and almost every other vendor in the data warehousing market is on the verge of announcing its own Internet strategy. The question is whether there's value in hooking up a data warehouse to the Web. The answer: the compelling advantages in using the Web for access to a data warehouse are almost the same as those for any other application:
Web access to a data warehouse is an open solution, allowing the use of any Web browser, such as Netscape Navigator, Spyglass Mosaic, Microsoft Internet Explorer, CompuServe, and Prodigy or Web servers like Netscape's Communication and Commerce servers, O'Reilly & Associates' Website, and Microsoft's Internet Information Server. Obviously, the look and feel of the client presentation can be enhanced with Java applets to build more powerful applications and utilize Netscape's JavaScript language. Any Web authoring tool can be used for the page design. As a bonus, Web browsers incorporate as standard features many attributes that are especially useful in analytical applications:
Good And Bad There is no question that Web access offers some clear advantages over existing client-server architectures-but there are likewise some very clear disadvantages. There is general agreement that the concept of a universal client container, indifferent to the physical medium in which it is conveyed, is a terrific idea whose time has come. After all, the online services and Internet service providers have blazed a trail by creating nearly ubiquitous access for more than 10 million people in the U.S. alone-a number that is still growing dramatically. Anyone connected to the Internet is at least theoretically connected to everyone else. With the rapid growth of the Web, and especially the emergence of de facto standards, such as Navigator and Java, everything looks pretty rosy. What could possibly be the downside? Six months ago, almost every corporate IT honcho had the same pat answer-lack of security. The thought of sensitive company information out there on the wild, unruly, and most-definitely Generation X-ish Internet was too scary. Encryption schemes and secure servers have allayed some of these concerns, but the phenomenon that settled the issue once and for all may be the killer app of the decade-the intranet. Security was easily handled by using all of the architecture, software, and applications of the Web and putting them behind a firewall, secure from the prying eyes of uninvited guests. Another big unanswered question was exactly how to get the corporate databases to talk to the Web. The CGI (Common Gateway Interface) protocols for message passing to and from Web servers require writing code in C, a decidedly retro approach in downsized IT shops. Another problem with using the Web to access data warehouses is that the interaction between a Web browser and a Web server is stateless, meaning that the familiar client-server notion of a connection to a data source just doesn't exist in the Web. The Web server acts as a message-passing server, responding to messages it receives, and in turn, contacting other resources on the network or replying to the Web browser. If each query to a database through a Web browser required a connection and login to the database, the accumulated overhead would have a withering effect on the database server and consequently on the user. The solution: Let the data warehousing software vendors do the hard work, building the Web gateways using the CGI API. Here's how the Web gateways for data warehouses work: When the user generates an OLAP query, product-specific tags are inserted into HTML and passed to the Web server. These are then passed to the Web gateway, which generates the request in the API of the OLAP server. The process is reversed on the way back to the user, passing the returned data to the Web gateway, which converts it to HTML and sends it to the Web server.
Upcoming OLAP Products For Web Access Essbase Web Gateway WebOLAP DSS Web Vendor Arbor Software Information Advantage MicroStrategy Sunnyvale, Calif. Minneapolis, Minn. Vienna, Va. 408-727-5800 612-820-0702 703-848-8600 Web site http://www.arborsoft.com http://infoadvan.com http://www.strategy.com Price $10,000 per Essbase server Starts at $29,995 Starts at $50,000 for DSS Web and DSS Server bundle Availability Second half, '96 June 1 June 1 DATA: VENDORS
By maintaining the connection to the data sources from the server and keeping track of each client messaging in from the Web browsers, the OLAP analytic engines control the whole interaction, which gets around the stateless problem of the Web. Obviously, since there is no way to poll the client to determine if it is still alive, there has to be some sort of timeout process, meaning that connections for a given thread are dropped after a certain period of inactivity. All these problems seem to be solved in the three products discussed below. But there's more. In OLAP tools, a common manipulation called drill down creates a series of analyses, each containing more detail about all or part of the previous analysis. This is often displayed as a series of overlapped windows in step fashion, lending a handy visual metaphor to the process. This memory of previous analyses is often referred to as a context of the interaction. Reproducing a facsimile of this in a Web browser requires some sleight-of-hand. Since the browser displays precisely what is requested in the HTML transmitted by the Web server, rather than updating or refreshing an existing page, the Web gateways for these products will have to produce much more complex pages that look like overlapped windows, something that, at this point, is much easier to do in a client-based graphical user interface. This will surely have performance implications as well. What about charts? Data visualization is a key element of analytical processing. The standard method for displaying graphics with Web browsers now is to broadcast the images from the server. Surely no one will tolerate the delay waiting for a 200-Kbyte GIF file to transfer, especially at 28.8-Kbps modem speeds. In order to draw charts dynamically at the client, based on a small burst of data from the server, additional capabilities need to be built into the products, most likely involving charting applets developed in Java or other emerging technologies, like Microsoft's ActiveX. Three Web gateway products are near release from Arbor Software, Information Advantage, and MicroStrategy. These contain only a small subset of the client-side features currently available in the vendors' client-side tools, but each can deliver at least some useful functionality in the near future. Describing these tools in a great deal of detail is not really useful, since they will quickly evolve and mature. Instead, I'll keep the descriptions brief and position the Web access features within the context of each tool's broader application architecture. Arbor Essbase Web Arbor's offering is perhaps the most ambitious of the early Web products. It includes not only OLAP manipulations, such as drill up, down, and across, pivot, slice and dice, and fixed and dynamic reporting, but also data entry, including full multiuser concurrent write capabilities-a feature that differentiates it from the others. Aggressively pricing Essbase Web at $10,000 per database server (not Web server), Arbor clearly sees the tool as a way to provide Essbase access to the masses at little marginal cost. Since Arbor sells Essbase only as a server, it does not have a client package that might be cannibalized by sales of its Web gateway product. Thus, Essbase Web makes sense from a business perspective. The Web product does not replace administrative and development modules, only user access for query and update. Information Advantage WebOLAP Information Advantage uses a server-centric messaging architecture, which is composed of a powerful analytical engine that generates SQL to pull data from relational databases, manipulates the results, and transfers the results to a client. Since all of the intelligence of the product is in the server, implementing WebOLAP to provide a Web-based client was straightforward. The architecture of Information Advantage's Web product is similar to Essbase's, with a Web gateway between the Web server and the analytical engine-although in this case, the data store and the analytical engine are separate, where Essbase is both a data store and an analytical engine. The first release of the product does not support the range of manipulations that Arbor's Essbase Web does. MicroStrategy DSS Web MicroStrategy's flagship product, DSS Agent, was originally a Windows-only tool, but MicroStrategy has smoothly made the transition, first with an NT-based server product, and now as one of the first OLAP tools to have a Web-access product. DSS Agent, in concert with the complement of MicroStrategy's product suite-DSS Server relational OLAP server, DSS Architect data modeling tool, and DSS Executive design tool for building executive information systems-generates SQL dynamically and relies on the relational database server to perform complex analysis, rather than creating a "cube" like most of the other tools. By inserting a Web gateway between the Web server and the DSS Server engine, MicroStrategy is able to replace the interactive DSS Agent front end with a Web browser, which passes requests to the DSS Server's API. Unlike Arbor, MicroStrategy sells a client product suite whose sales may be reduced when companies license the Web product. The client product suite offers a wide range of capabilities that will probably not be duplicated, at least not right away, in the Web product. One such feature is the metadata manager, DSS Architect, as well as development tools for building complex templates (arrangements of dimensions and facts), metrics (raw or derived facts), and filters. It seems clear that all data warehousing vendors will pay lip service to the Web, with their efforts consistent with the approaches taken by the three pioneering vendors. Server-centric tools, such as Information Advantage and Prodea Beacon from Platinum Technology, and the multidimensional databases-Essbase, Oracle Express, Planning Sciences' Gentia, Kenan Technologies' Acumate ES, Holistic Systems' Holos, and Pilot Software's Pilot Server-appear to be in an excellent position to take advantage of their architectures to provide easy access from a Web browser. Surprisingly, many of the server-centric tools have yet to announce such a product. As to why the others have not done the same, we can only speculate. Apparently, having the analytical engine as a middle tier between the client and the database server is not sufficient to quickly take advantage of the Web. I suspect many of these tools, built around proprietary messaging schemes and lacking open APIs, are having some difficulties accommodating the stateless nature of the Web. Client-centric tools, such as Business Objects, Software AG's Esperant, Cognos' PowerPlay, or Brio Technologies' Brio Query, may be in a tough spot if the Web becomes a preferred approach. Since they rely on the client-particularly a Windows or Mac machine-loaded with disk space, memory, and processing power, to perform a substantial amount of the analysis, the Web doesn't provide much relief from supporting a fat client. In order to take advantage of the architecture, these vendors will have to port their engines to a server in addition to developing the Web gateways. Another client-side OLAP tool vendor, Select Star in New York, is taking a different approach. In its as-yet unnamed offering due out in a few months, Select Star opted to provide a more general-purpose toolkit for Web development. Though the toolkit initially will rely on its client-based OLAP engine, StarTrieve, Select Star is targeting a much wider audience than just OLAP. Features Now And Later When you evaluate these products, please keep in mind that the features you see now should not be the deciding factor. More important is choosing a tool that fits the architecture of your data warehouse, since you will want an architected solution. Second, buy the company and not the product. It's very early in the development stage and the features included in the first release of the products are not an indication of the quality of one vendor versus another. What is clear is that the Web is here to stay and can play a valuable role in data warehousing, even in the short term. Obviously, existing client-server approaches will not be displaced anytime soon.
Neil Raden is president of Hired Brains, a consulting firm specializing in data warehousing based in Santa Barbara, Calif., Chicago, Toronto, and New York. He can be reached at 805-886-8892 or nraden@hiredbrains.com InformationWeek http://techweb.cmp.com/iwk |
||||||||
Copyright © Hired Brains Incorporated,
2002. All rights reserved. |
