Hip hip hooray, the first object-oriented databases are rolling off the production lines courtesy of companies such as Burlington, Masachusetts-based Ontologic Inc (CI No 1,302) and Versant Object Technology Corp of Menlo Park, California (CI No 1,435). But aside from employing one of the most fashionable software techniques around, what does the arrival of object-oriented databases mean to the computer industry. Are there grounds for excitement? According to Robert Martin, a vice-president at Ontologic, an object database integrates an object programming language (usually C++) with database functionality, thereby fulfilling the requirements to build high-performance database applications with well-structured, easily maintained applications. Such databases save objects on the disk so that they can be used in future sessions and shared by many users, and they also provide applications programs with an abstract interface to data, enabling programmers to change or reorganise the database without changing any application code. Martin claims that object databases have one major advantage over relational databases: namely that relational databases aren’t flexible as they have to store all data in fixed-length records – all applications data must be converted to these records making relational databases slow and prone to error.
Abstract pointers
However, with Ontologic’s product Ontos, the database can directly store and retrieve objects as a whole rather than trying to fit every thing into records as a relational system does. Furthermore, Ontos uses abstract pointers to implement relationships among objects whereas relational systems use keys to join operations, and while a relational system cannot get hold of an entire complex object in one query, an object database can retrieve an object composed of many other objects as a whole. As far as Martin is concerned, this means that object databases have three highly marketable features vis a vis relational databases: they are designed to handle any type of data (rows and columns, graphics, finite elements, bit-map imaging and so on); secondly, they can handle such complex data at high speed, being up to 1,000 times faster at graphics and documentation than relational systems; thirdly, they offer high programmer productivity, since applications are easy to build and maintain using them. So far all this might sound like little more than vendor-hype, however, the politics of database marketing has recently been stirred up by the publication of the ButlerBloor report Database: An Evaluation & Comparison (CI No 1,438).
In the report the authors Martin Butler, Robin Bloor and Paul Beach make some striking statements about relational databases. They say that they found no evidence to suggest that many relational database management systems can offer satisfactory solutions for the large, complex database implementations that are needed for an organisation’s core business data processing.
By Katy Ring
The software triad continues by stating: performance limitations associated with the relational approach may cause a backlash in the market, with the re-emergence of non-relational options… object-oriented methods… will come to dominate the database market over the next three to five years. Messrs Butler, Bloor and Beach are particularly unhappy about Oracle’s dominance in the database market saying despite its dynamic marketing, many of Oracle’s products have been relatively static and in the database area it has been overtaken in sophistication by Ingres and Sybase. If it loses its sales momentum it will find it difficult to recover. On the next page the Canadian company Cognos is discussed as follows: PowerHouse StarBase is an impressive product with a number of unique features, such as multiversioning, and could become quite popular if it acquires sufficient momentum. While both these statements are very probably true, each contains an extremely large if qualification which centres around the marketing of databases. ButlerBloor Ltd finds the concept of hard sell d
istasteful because it draws attention away from the technical merits of a product, which is a perfectly reasonable viewpoint, except when it clouds the issue of forecasting the development of a market. Will object-oriented databases really dominate the market in three to five years time? And are relational databases about to die out and with them a huge company such as Oracle? From talking to Ontologic’s Robert Martin the answer to both these questions would seem to be no. To take the second question first. Most relational database vendors are beginning to add object-oriented features to their relational databases – this is only at the front end, but with time and money there is nothing preventing them moving to object databases. The issue is more of a religious one than a practical one – those vendors that are adamant that the relational model rules, despite a changing market, may suffer. However, ironically enough for the ButlerBloor scenario, top of the list of relational vendors that may go under in this way are the technically led companies that will, arguably, find shifting to another database paradigm more difficult than market-led companies. As for the first question, although Martin describes his company as precocious, it is largely marketing its Ontos database for the benefit of customers that never moved to a relational database in the first place. In other words, Ontologic’s natural constituency is the computer-aided design and manufacturing community that couldn’t have used a relational database because its application files are too complex. They are also confirmed Unix users who are familiar with the C language. Where Martin does replace a relational database product it is more than likely to be Sybase – a fact to be weighed when judging which relational vendors may fall prey to rival object databases. Martin argues that object databases will take hold of the database market much faster than relational systems did in the 1980s.
Content from our partners
Rising wave
This is because they already have a standard language interface in C++, they are in a position to ride the rising wave of network computing because they can take advantage of graphical applications, and if, like Ontos, they support SQL then they also provide a link with past applications and data. However, when it comes to the business community where the majority of software is sold, object databases don’t look as if they’re about to stage quite the revolution they could in the design world. Although Martin argues that people are clamouring for complex graphics in their management information systems, and that banks, stock exchanges and leading high street chain stores are all in the process of evaluating object-oriented technology, he admits that object databases are no faster than relational databases when it comes to things like on-line transaction processing. The one factor that may unleash object databases onto the business world is the arrival (expected sometime next year) and consequent success of object-oriented Cobol. But since no such language has yet appeared – and when it does it won’t be compatible with previous versions of Cobol (CI No 1,427) – the speed with which object databases penetrate the business world is likely to be slow.
