Is database technology a commodity software or vital piece of enterprise infrastructure? Until recently the giants of the database industry would have insisted that the latter definition was correct. But now the winds of change are blowing – it may only be a breeze at the moment, but there is the prospect of a hurricane in the not too distant future. There has been a subtle change of stance by the database market leaders, not a Pauline conversion as yet, but definitely a shift to keep their options open.
It is understandable that this should be the case. Lessons have been learned from the fate of John Cullinane, CEO of hierarchical database firm Cullinet. Cullinane was the Larry Ellison of his time, the thought and market leader in the database space. But he refused to accept the coming of the relational database model and like King Canute, tried to turn back the tide. Instead he and his company drowned while Ellison and Oracle stepped up to take their place.
Seeing the red flags
So it is hardly surprising then that Oracle should be one of the firms that (a) will talk of the day when database technology becomes a commodity to be given away and (b) will embrace the idea of a new model that threatens to become a potent force: the open source database.
There is a stirring in the relational tribe. IBM now offers a lower-end version of its DB2 product for free, following similar moves by Microsoft and Oracle. At the same time, companies such as Ingres and EnterpriseDB are ramping up high-end database offerings based on open source.
"The way Oracle licenses does not allow you to follow your software lifecycle. You are not able to reduce or add support as you need"
David Manifold, director of database services, Sony Online Entertainment
Of course, the world has been here before. Open source has famously been about to break into the mainstream on many occasions. Sun Microsystems’ OpenOffice ought to have crippled Microsoft Office’s installed base by now, if Scott McNealy’s overenthusiastic hyperbole is to be believed. But on the other hand, Linux is now regarded as a viable enterprise operating platform for large corporates, so clearly it is just a matter of the ‘right time, right place’ for open source technology.
So what are the prospects for open source databases? Well, for starters even if their revenue run rates are small compared to the likes of Oracle and IBM, they are currently racking up the highest growth rate in the database market, according to analyst firm Gartner. “The combined category of open source database management systems vendors, which includes MySQL and Ingres, showed the strongest growth, although it was one of the smallest revenue bases,” says Colleen Graham, a principal analyst at Gartner.
“These open source products continue to improve in functionality and scalability – and database tool vendors are beginning to provide support for these offerings.”
Overall revenues in the database market grew by 8.3 per cent in 2005 – the latest available figures – to $13.8 billion. Oracle remained the largest vendor with a 48.9 per cent market share, followed by IBM at 22 per cent.
Microsoft was a fair bit behind with 15 per cent while Teradata and Sybase took 3.2 per cent and 2.9 per cent, respectively. The remaining 8.2 per cent went to smaller vendors and open source database providers. But Linux grew the fastest of all the database platforms at 84 per cent – a fact that has not gone unnoticed by the likes of Oracle and IBM. Both have their own open source strategies in place.
Oracle had previously bought two small open source database companies – Sleepycat and InnoDB – but upped the stakes last year with a futile attempt to buy MySQL. The very fact that it tried is indicative of the changes that the open source movement is likely to have on the established relational players.
“Alternatives such as MySQL are attracting a good deal of attention and loyalty from a new generation of database developers,” says Carl Olofson of International Data Corporation (IDC). “IDC believes these open source vendors could ultimately spur a fundamental change in the way that RDBMS products are priced and licensed.”
There are already enterprises that have made the leap into wide-scale deployment of open source database technology – typically as replacements for relational incumbents.
Among them is Sony Online Entertainment, which operates an online gaming network. It is replacing its Oracle databases with EnterpriseDB, initially for back-office applications such as billing systems, but moving into front-end customer facing applications as well as basing new games development on open source platforms. The firm develops a number of popular massively multiplayer online role-playing games, including EverQuest and Star Wars Galaxies.
"The primary value from open source is to take things from within the infrastructure that are commoditised in order to release value at the business edge"
JP Rangaswami, CIO, BT Global Services
With hundreds of thousands of subscribers to service, the company has the challenging database requirements to store and manage terabytes of data about the virtual ‘worlds’ in which game characters live, battle and fall in love. For example, as part of the Star Wars Galaxies game, players can customise lightsabers, X-Wing fighters and blaster guns to their particular specifications.
This means that every object in the game must be stored individually, rather than storing a generic version for every player.
Licence to print money
The move to an open source database was driven initially by a need to cut software licensing costs. The sheer complexity of its licensing policies has always been a weak chink in Oracle’s corporate armour. “Our primary business is building massively multi-player games,” says David Manifold, director of database services.
“We’ve realised over time that games like Star Wars Galaxies can best use a database as part of their architecture. It allows us to host hundreds of thousands of global online gamers. Until now, we used the Oracle database and had become very dependent on it. We realised that the cost of the Oracle database was the most expensive element in many of our systems. We were using less expensive hardware and then we’d get the bill from Oracle to deploy and we would be in shock,” says Manifold.
“The licensing was very inflexible. If anyone has had to deal with Oracle licensing they know what I’m talking about. The way Oracle licenses does not allow you to follow your software lifecycle. You are not able to reduce or add support as you need. There’s always a demand for new ‹ databases. It’s a constant battle to manage your Oracle licences and make sure you’re compliant.”
Cost benefit analysis
Sony Online Entertainment expects to save “a substantial amount of money” as a result of its decision. Oracle’s database has a list price of $40,000 per processor for the Enterprise Edition. EnterpriseDB charges an annual subscription of $5,000 for its database with the top-end platinum support package. Sony Online Entertainment is typical of the way open source databases will most likely penetrate the mainstream – coexisting with installed relational technology but increasingly deployed for new applications instead of the incumbent provider.
Another driver for choosing EnterpriseDB was that while it is based on PostgreSQL – an open source relational database management system developed by a global team of volunteers – it offers a high level of Oracle compatibility.
This meant that as much as 90 per cent of existing Oracle-based applications do not need to be rewritten to run on EnterpriseDB while Oracle-trained staff can immediately work comfortably and efficiently with the open source-based alternative.
“We had built up a very big pool of Oracle knowledge inside Sony and we didn’t want to throw that out. We couldn’t afford to spend too much time on a migration especially as we had so many Oracle applications that we’d developed inhouse,” says Manifold.
The move to open source databases was not entirely a leap of faith. Sony Online Entertainment had already successfully deployed Linux, Tomcat, Apache and Hibernate applications.
“We had migrated away from proprietary Unix to Linux systems,” says Milford. “We find it to be very stable with good uptime. It provides us with cost savings and stability.”
For the future, Sony Online Entertainment is looking at the prospect of moving its mega-popular game, The Matrix Online, off Oracle to EnterpriseDB as well as planning to deploy a further three new games on the open source platform by 2008. For each game deployed, it expects to see database total cost of ownership saving of about 80 per cent when measured against current costs.
The price of freedom
Other firms take a slightly different tack. JP Rangaswami was CIO at investment bank Dresdner Kleinwort for five years before becoming CIO at telecommunications giant BT Global Services in October 2006.
While at the bank, Rangaswami says he was “given the freedom to experiment with real innovation in technology and its application”. That included open source database technology and it is something that he has brought with him to the company. While recognising the cost benefits of an open source database model, Rangaswami is keen to make the point that users need to be wary of certain terms being bandied around.
“It’s free as in freedom, not free as in ‘gratis’,” he points out. “Some people are going to make money out of this.”
Open source appeal
So if cost is not the main driver, what is the motivation for deploying an open source database such as MySQL at a firm like BT Global Services?
“We are transforming from being a product shop to one that is looking at functionality and best fit with what our customers want,” he says.
“The primary value from open source is to take things from within the infrastructure that are commoditised in order to release value at the business edge. We had to wait for technology to be made available at the enterprise level. My experience is that there is a shift from individual to enterprise capacity. It used to happen the other way around – enterprises were first adopters and it then migrated to the low end service.
“Now we see technology embedded first at an individual or personal level, then it moves into something that is suitable for scaling across an enterprise. You can see the same shift in something like Skype which was not seen as an enterprise offering in its first versions.”
Of course, there will always be dissenting voices, those who are dyed-in-the-wool Oracle or IBM advocates. Rangaswami concedes that there will be different degrees of receptiveness to a new model. “You have the early adopters who are happy to be first; then you have the large middle group who are open to change but look pragmatically at their options on a ‘horses for courses’ basis; then you have a small group who say ‘over my dead body’.”
But he concludes that open source database technology is here, it is enterprise useable and people better get used to the idea – users and vendors alike. “Today it would be hard to find a large organisation that doesn’t have Linux in some form,” he argues.
On the horizon
“MySQL is probably in a lot of organisations that don’t even know it’s there,” says Rangaswami. “The shift to open source has occurred in such a way that we’re not talking about labels but about its value.
“It would be hard for someone to come up with a list of mainstream IT criteria that didn’t have open source on it. As a CIO, if you’re not looking at open source, you should be.”
On the other hand, some still argue that open source databases are years away from offering a viable, enterprise-ready alternative to the likes of Oracle and IBM, including the outsourced thinkers of the business world. The capabilities of open source databases are nowhere near the commercial players currently. In fact, it could take as long as a decade before open source databases can meet the business intelligence and data warehousing demands of most large enterprises.
One reason for this is that most open source development effort still goes into contributing code for Linux evolution. The community for open source operating systems is entirely different to open source databases.
So while users experiment and pilot open source database deployments, there may still be time for the existing relational kingpins to get their act together and avoid the fate of John Cullinane. The alternative is that they will all be washed away like that other silly old Canute.