From Multimedia Futures, a sister publication
Database firms are squabbling over their ability to support multimedia, an enormous business opportunity for the database vendors. Take the Web – its key selling point is that it offers access to many sources of information and many types of data – text, audio, video, graphics, still images and sound. But how can these volumes of datatypes be stored and handled? The datatypes and relationships are beyond the technical capabilities of relational databases. RDBMS do not have the architecture, storage or query mechanisms to exploit new information and multimedia sources. Now, established vendors are incorporating object technology in their databases as a means of dealing with the new data types. Objects have lurked around the IT subconscious for years, but object technology databases remain pretty esoteric.
By Janice McGinn
Few IT managers are brave enough to adopt them as business- critical database technology, although there are exceptions. The finance industry benefits from object databases’ handling of time series data like international currency movements and exchange rates. The way they handle relationships between multiple data types is far superior to relational technology. An RDBMS can churn through high volume simple queries, but for complex datatypes, the RDBMS incurs significant performance penalties and offers few sophisticated query mechanisms But there are good technical and business reasons why pure object databases have not been adopted by the mainstream and the trend is for the major suppliers to incorporate objects databases rather than replace RDBMS. The debate about database technology for the multimedia is not about objects versus relational versus flat file. The major object database vendors agree that object databases have to integrate with relational technology. The key requirement is twofold: to protect user investment in RDBMS; and to provide new multimedia capabilities. Two approaches to multimedia databases are dominating the debate. The first is a hybrid technology, the object-relational (ORDBMS) models supported by Oracle and Informix. A second approach comes from the object database camp, led by Object Design.
Objects everywhere? Not quite
Despite a period of marked financial instability in the early 1990s, Object Design now holds the largest share of the ODBMS market, according to IDC. Nonetheless, the company’s president, Bob Goldman, is pragmatic about its place in an organization: We don’t expect anyone to dump Oracle, he says. But business needs an environment that merges relational and multimedia datatypes. Think of ObjectStore as a data warehouse for old and new technologies. The company announced its Web and multimedia strategy last week. It says that the new suite will enable users to build dynamic Web sites and applications. A new tool, ObjectForms, is a development environment for Internet and Intranet applications; and a suite of six Object Managers provides native support for text, audio, video, image, HTML data and Java applets. The data model support objects, extended data types and their relationships, and to create dynamic Web content, the suite allows users to retrieve objects, search and assemble them instantly on pages, and then deliver the pages quickly to a Web site. Despite advocating coexistence with RDBMS technology, European vice president, Jim Beagle, says it is evident by the strategies of the relational database vendors, that RDBMSs are not suited to a data-rich environment. He argues that object technology is a natural fit. It enables users to build complex web sites with optimum performance. ObjectStore’s architecture manages extended relationships better than the traditional databases that require data to be stored in tables and rows. ObjectForms and the Extended Object Management Suite are aimed primarily at Web application developers and supporting multimedia data types. Specifically, the Forms tool supports HTML page generation, forms processing, sca
leable architecture options and connectivity between ObjectStore and HTTP Web servers. It includes tags for extending HTML files, and reads these tags to determine how to query or update the database, then creates HTML to display the results. The Extended Object Management Suite supports extended data types. Goldman says users can store, query, manipulate and reuse extended data types, and, supposedly, eliminate the need for developers to build their own libraries of objects. The Image Object Manager includes a query- by-content engine developed by Virage Inc. It allows users to target an image, and search for all similar images – in online clothing catalogues, for example. The Text Object Manager uses Verity Inc’s , which provides automatic indexing and very high- speed searches. That’s one viable approach to multimedia and the management of Web data. The relational vendors are taking a less considered approach. They are ramping up their product lines for multimedia and Web developments. But in ill-considered marketing programs, instead of focusing on the benefits of multimedia- capable databases, Oracle and Informix are exhibiting the debating skills of the kindergarten. Briefly, both companies have reiterated their commitment to the Web and multimedia technologies. Both emphasize their ability to handle different datatypes and claim that, via their Universal Servers, users will be able to store and access diverse datatypes. Oracle Corp has market share to spare. It dominates the user base by a combination of reliability, habitual buying and muscular marketing. In theory, it should be targeting this new market with confidence. But despite launching its Universal Server, Oracle version 7.3, only last month, Oracle is not ebullient. It is defensive, and stands accused of delivering an interim product with no object-relational capabilities. 7.3 is a token product, a mere nod in the direction of the Web, say the critics. They argue that users depending on their longtime RDBMS supplier to take them into the world of multimedia will be disappointed with version 7.3. For example, the Meta Group slated it, saying that 7.3 is a new media assortment that combines pre-existing and separate, Web, text management, messaging and multimedia information management products….under a single marketing umbrella. It does not provide a totally integrated approach to new media data management.
Content from our partners
7.3 not an interim, says Oracle
Oracle acknowledges that 7.3 is not truly object-relational. It denies that it is an interim model, but admits that fuller functionality won’t come until Version 8. It also denies accusations from Informix that Version 8 was due to be announced and shipped this quarter, saying that it will be announced and ship, as planned, during mid-1997. Well, maybe it did or didn’t commit to shipping during 1996. What matters is that there was an expectation which Oracle did not dampen until too late. So are there any serious implications for users? If an organization plans to exploit object-relational database capability, with Oracle as the key environment, the strategy will have to be postponed. If users are prepared to swap suppliers, Informix is banging a loud drum for its recent acquisition, Illustra Information and the object-oriented datablade modules.Informix is garnering plaudits for Illustra’s plug-in datablades. Bloor Research, a reliable commentator on database technology, has said that optimization of data types and speed of retrieval are two areas for which Illustra was designed, and that Illustra’s capabilities are beyond those added by Oracle for Oracle 7.3. Andy Bailey, Oracle UK’s marketing manager, argues that the Informix-Illustra technology has not been merged yet. He says that Informix has promised to convert code paths… they will deliver something, but not true single source convergence. He claims that most analysts don’t believe it can be done within the stated timescales. So sorry, Mr Bailey. But most analysts seem to believe Informix. It plans to rollout datablades thro
ughout 1996, intending to have 100 by the end of this year, 1,000 by the end of 1997. Every datablade released can be customized by users and incorporated into its Universal Server, says Informix. For the record, Informix states in its Universal Server product plan, that as of January 1996, the release schedule is expected to be: Universal Server Datablade Developer’s kit in second quarter 1996; Illustra to Informix Gateway in second quarter 1996; Universal Server in the fourth quarter 1996. Illustra’s datablade modules are recognized by competitors as good multimedia handling technology. Informix biggest strength, says Object Design. But Informix is not content. It launched a spoiling campaign, claiming to expose the technical and business shortcomings of Oracle 7.3. Its PR machine leaked criticisms and comments shortly before Oracle announced 7.3, and Michael Stonebraker, Informix recently-acquired guru, let loose a stream of vitriol during a recent press tour (He co- founded Ingres Corp, and then established Illustra Information.) He accuses Oracle of killing Ingres, and of casting a fog over its own products and those of its competitors. Informix code has been rewritten from scratch, unlike Oracle’s old and tired coding. ObjectStore’s Goldberg says that, clean code or not, the problem facing both is that their fundamental architectures are tabular, unsuitable for storing multimedia types like images and sound, and for establishing dynamic binding relationships. For example, an analyst might want information on a company’s financial performance, (text and numbers), along with any related sound bites (from television or radio) and video clips (from any number of internal and external sources). He wants a complete financial profile without having to join up tables manually in a relational structure, says Goldberg. Oracle’s Version 8 is expected to support video and audio integration with full text search, parallel index construction, bit-mapped index technology, and a front end with multi-dimensional data presentation. Object- orientation will be stronger and there will be SQL3 support and user-defined datatypes. Informix claims to have two-year technical lead on Oracle. But if the world’s most successful database company delivers its promised multimedia functionality, Informix will need more than a spoiling campaign to maintain its headstart. Neither company, despite the dust and furore, has a full-function Universal Server on the market. Few users want to change database suppliers, and until they are in a position to make valid comparison, it is unlikely that Informix can persuade them to swap allegiances. Analysts judge. But the jury is out on ORDBMS.
