Dave Barry once said, “Buying the right computer and getting it to work properly is no more complicated than building a nuclear reactor from wristwatch parts in a darkened room using only your teeth.”

Barry was referring to a home computer, so imagine the pressure on a newspaper IT director charged with buying an enterprise-wide publishing or business system, which needs to be robust, reliable, highly scalable, secure, and fully available 7 days a week, 24 hours a day and 365 days per year.

 

For the past several years, most companies in the media industry have continued to rely on relational database platforms like Oracle or Microsoft SQL Server to power their business-critical software systems. These databases provide the optimum technology foundation for the complex, deadline-driven, multimedia environments in which our customers operate.

Recently, however, a few industry suppliers have introduced editorial content management systems based on object-oriented database platforms from companies such as Versant, InterSystems (Caché), Objectivity and GemStone.

Proponents of object-oriented databases often cite the following three main advantages over relational database technologies:

•They run faster for transactional applications.

•They offer better developer productivity.

•They are easier to manage.

Limitations abound

It is true that object-oriented databases are designed to make development easier. However, these platforms are lacking in functionality, reliability and media industry experience.

In response, some industry vendors, such as Atex, have taken a “best of both worlds” approach, engineering object-relational mapping tools to provide customers with the rapid response of object-oriented development while allowing the back-end to remain on robust Oracle or SQL Server database platforms.

Yet the fact remains that databases such as Oracle and SQL Server are superior to smaller, lesser well-known object-oriented database (OODB) architectures.

Here’s why:

•Oracle and SQL Server databases deliver consistently better performance than their OODB counterparts.

On March 10, 2008, the Transaction Processing Council (TPC), an independent non-profit software testing organization, conducted its latest performance benchmarking tests for databases ranging in size from 3TB to 30TB. Oracle ranked No. 1 in performance in all these database size categories. In addition, Oracle databases led all competitors in enterprise software benchmark tests of both clustered and non-clustered environments.

TPC also recently completed a benchmarking test of Web-based applications running on a variety of hardware platforms, including those from IBM, Hewlett Packard, Fujitsu, and NEC. Each hardware/software combination was evaluated in terms of online processing performance. Oracle and SQL Server databases powered eight of the top 10 configurations tested. The other two leading performers ran on IBM’s DB2, another relational database. Not a single OODB configuration shows up in the top-10 list.

One of the main reasons that OODBs often perform poorly is that many OODB implementations do not provide the kind of sophisticated data-locking that Oracle and SQL Server offer.

Without these built-in tools for row-level locking, application software running on an OODB can often lock huge chunks of data for every transaction that’s processed. The problem is compounded in sophisticated applications —such as content management — where multiple relationships exist among data objects and database transactions can often lock more than one page in more than one table. As a result, multiple users competing for database requests often have to wait until one or more pages of data is unlocked before the transaction can be completed. This translates into longer wait times for each user on an OODB system.

By contrast, by only locking a single row of data for each transaction, granular locking schemes employed by Oracle and SQL Server provide the optimal mix of secure data management, load-balanced support for many thousands of concurrent database requests and rapid end-user response times.

•Oracle and SQL Server provide better, more flexible tools for querying and accessing content in the database.

One of the hallmarks of Oracle and SQL Server databases is their ability to construct sophisticated queries to access and report on content and metadata in the database. Queries can be created and run by end-users in a highly structured manner, and frequently-used queries can be optimized for performance through the use of stored procedures. In addition, Oracle and SQL Server provide numerous built-in tools for constructing and executing ad-hoc queries.

OODBs, on the other hand, typically provide much weaker support for ad-hoc queries against the database. Navigation through a complex database can be challenging with an OODB and query optimization tools and functionality lag behind those available in major relational database systems. This leads to a higher IT cost when integrating OODBs into existing and future enterprise infrastructures, such as Web portals, digital archives and other systems demanding flexible workflow tools.

The InterSystems Caché OODB, for example, was developed in the 1970s for processing hospital patient records. Caché’s original name was MUMPS, and it was designed to handle a large volume of relatively small transactional records with few interrelationships with other records, such as the type of data that would be found in patient information cards. Such applications are vastly different from the multithreaded relationships, dependencies and conditionals required in a newsroom content management application.

•Oracle and SQL Server provide a more secure database environment for business-critical applications requiring 99.999 percent uptime.

The U.S. National Computer Security Center (NCSC) has given Microsoft SQL Server a C2 Certification, which is the highest government security classification available. In addition, Oracle is consistently the market leader in formal security evaluations with 19 established and validated security certificates from independent industry global testing agencies (see http://www.oracle.com/ technology/deploy/security/seceval/security-evaluations.html for details).

•There are a lot more trained resources available to support and maintain Oracle and SQL Server databases than any OODB on the market today.

As of March 2008, Careerbuilder.com has more than 4,500 resumes in its database with candidates who have Oracle database experience. An additional 4,000+ candidates have Microsoft SQL Server experience. By comparison, only 147 CareerBuilder candidates have Versant OODB experience and only 61 have Caché experience.

Matching qualified job candidates with open positions remains a major challenge for technology companies. The North American high-tech job site Dice.com illustrates the wide disparity in resumes and available jobs for each database platform: Oracle (19,813); SQL Server (14,702); Versant (3); Caché (20).

Furthermore, finding individuals with proven, deep experience in OODB administration will be challenging. Granted, many OODB users feel that OODBs require less administration in the early stages, but like most growing systems OODBs still require tuning in preparation for deployment. Switching from a relational to an OO database paradigm also will require some training and mentoring because issues such as performance, locks and joins often require different approaches.”

Finally, the difficulty in finding trained resources to support a Versant or Caché database is compounded for companies considering a move to a hosted or managed services model.

Most of the major data center providers (e.g. NaviSite, Rackspace, Savvis, Qwest, etc.) are fully capable of providing first and second-tier support for Oracle and SQL Server databases. Providing similar support for lesser-known OODB platforms would add significant cost to the monthly hosting fees because of the extra training and specialized skill-sets required.

Consequently, even though the initial price tag of an OODB software license may be lower than that of an Oracle or SQL Server license, the total cost of ownership (TCO) is often much greater. Because there are fewer experienced OODB specialists on the job market, training costs can be quite high when compared to Oracle or SQL Server experts, who are able to bring their training and experience to a new position, and effectively “hit the ground running.”

Lost flexibility

Clearly, OODB solutions like Versant and Caché are valuable for software developers who want to build object-oriented platforms where the database design is tightly coupled with the application. This tight coupling simplifies both the design and the coding process, and helps speed up development times. The downside of this simplistic approach is lost flexibility and software agility. With Oracle and SQL Server, the applications are loosely coupled, and there is a valuable abstraction layer sitting between the applications and the database.

The loosely coupled model offered by Oracle and SQL Server is ideal for the fast-changing, convergent, and agile technology foundation essential for high-powered publishing applications. With Oracle and SQL Server, database schema changes can be made independently of the actual applications.

Conversely, in an OODB, because the applications and the database are so closely linked, schema changes often involve a system-wide recompile. This translates into an OODB operating environment that, in many instances, is not responsive to changing business needs or new market opportunities. In addition, companies selecting an OODB solution often remain heavily reliant on their database or application vendor for any changes required, which further increases the TCO through ongoing year-over-year support and professional services expenditures.

Peter Marsh is chief technology officer of Atex Ltd. He can be reached at pmarsh@us.atex.com