Database wars revisited

In October 2004, at a conference for technology industry executives held in Santa Barbara, California, Oracle president Charles Phillips delivered a stinging indictment of the company’s open source rival in the database software market, MySQL. “We’re both in the transportation business,” he said. “We have a 747 and they have a Toyota.”

That Phillips would take such a blatant swipe is no surprise: Oracle is famously aggressive in dealing with any threat, real or perceived. During the so-called ‘database wars’ of the early 1990s, chief executive Larry Ellison is alleged to have shown top managers at the company a photograph of a competitor’s headquarters and told them: “We’re going to run them out of business and buy that building, which we’re going to bulldoze. After that, we’ll salt the earth. Then we’ll go after their families.”

That example comes from Karen Southwick’s biography of the Oracle CEO, tellingly entitled Everyone Else Must Fail. But according to Matt Aslett, an analyst at IT market research company The 451 Group, Charles Phillips’ comments demonstrated quite neatly the differing business strategies and selling points of Oracle and MySQL.

In fact, he says, his attack did that job so well that MySQL executives began regularly citing it themselves as evidence of the attractions of its low-cost approach. Almost five years on, with the announcement in April 2009 that Oracle is to acquire Sun Microsystems (which itself bought MySQL just four months earlier), industry speculation that Oracle will kill off the open source database product is rife.

Oracle, however, isn’t talking. When approached by Information Age, a company spokesperson said that Oracle executives were not prepared to comment on MySQL’s future at this stage.

While a slow, lingering death for MySQL is a possibility, Aslett believes such a move gives neither Oracle nor MySQL enough credit. “Oracle and Larry Ellison didn’t get where they are today by ignoring a business opportunity, and MySQL represents a more significant business opportunity for Oracle alive than dead,” he recently blogged.

For a start, he explains, MySQL’s strategy of avoiding competition with Oracle and IBM has enabled it to establish a substantial business under the noses of the database giants by targeting a market (web applications) that was not, at one time, of much interest to those giants. Now that this market has grown to represent a lucrative business opportunity, these companies are painfully aware that MySQL – and other open source database companies – have been successful precisely because they offer something different from established, and largely proprietary, database choices.

These companies (which include MySQL, Ingres and EnterpriseDB) have been so successful, in fact, that analysts at research company Gartner recently predicted that the open source database market will be worth $1 billion by 2013 – an impressive total for software that is essentially free, but which still manages to attract a healthy revenue stream in the form of ongoing support and maintenance charges.

In 2008, open source database use grew 50%, according to Gartner, with 73% of companies surveyed deploying them somewhere in their IT infrastructures, up from 49% in 2007.

Sales, meanwhile, grew 42% over the year. Such metrics have not gone unnoticed by Oracle, IBM and Microsoft, who together command an 88% share of the $27 billion worldwide database market.

And the opportunity has sparked a strategic battle between these vendors and their open source rivals that harks back to the days of the database wars of the 1990s. At the same time, this fractious situation makes choosing an open source option a tricky business for IT decision-makers.

Cost equations

There’s no doubt, however, that it’s a decision they’re more than willing to make – and the main reason, in a time of severe constraints on IT spending, is cost.

“In the current economic climate, the number of enquiries to Gartner about the use of open source databases in place of commercial databases is escalating rapidly,” says Gartner analyst Donald Feinberg. Licence cost and maintenance charges are the primary area of cost reduction that organisations achieve in taking the open source option, whether they’re looking to replace an existing, proprietary database or acquire a new one to support new applications.

In both scenarios, he points out, “The licence cost of the open source database is zero, compared with a range of costs for commercially licensed databases, from around $25,000 per socket to as much as $50,000 per core.”

Further, ongoing savings occur with maintenance costs, he adds. Maintenance costs for proprietary or ‘closed source’ databases, he says, amount to 20% to 22% of the original licence fee each year, including customer support and access to new releases.

Based on the list prices published by vendors, those charges can range from about $11,000 per server per year to as much as $44,000 for an eight-core server. By contrast, open source database subscription support on a 24/7 basis ranges from $5,000 to $10,000 annually for the same eight-core server. Over a period of five years, the company that deploys that server could expect to see savings of almost $370,000.

The economics, then, are compelling – but other factors need to be taken into account by prospective customers. Open source databases do come with some upfront, first time charges relating to IT staff training, management tools from third-party vendors and other software necessary for functions such as high availability, which may not be included with the new database.

In addition, most commercial database vendors offer substantial discounts, and some – Microsoft and Oracle, for example – have begun to offer unlimited licence agreements (ULAs) that can substantially change the cost dynamics.

They are also giving away cut-down versions of their products, for free. IBM, for example, offers DB2 Express-C, a free-to-download and redistribute version of its flagship DB2 database, launched in early 2006 following similar moves by Microsoft (with SQL Server 2005) and Oracle (with 10g Express Edition).

The DB2 Express-C database is not open source, explains Adam Jollans, open source strategy manager at IBM, “but it is free, which makes it a very good product for organisations that want DB2’s enterprise capabilities but can’t build a business case for the investment it would involve”.

Performance pressure

Cost issues aside, IBM, Microsoft and Oracle all insist that there are still many workload types where open source databases simply aren’t up to the job.

“Databases are complex pieces of software,” says Jollans at IBM. “Data integrity and optimisation are intrinsically hard things to do, and open source databases are lightweight in comparison with DB2. MySQL may be very good at supporting web front-ends, but it doesn’t offer the enterprise capabilities required for mission-critical, highvolume back-end systems.”

Gartner analyst Feinburg is inclined to agree: “We believe that, today, non-mission critical applications are the best choice for use with open source databases.” The exception, he says, is Ingres, which has been available for over 25 years and already has a broad base of organisations using it in mission-critical applications.

One such customer is the Irish government, which has run its core tax systems on Ingres for years, according to Emma McGrattan, senior vice president of engineering at the company.

In fact, she adds, executives at many organisations simply aren’t aware that their IT department has been deploying open source databases for years, and often to support high-volume workloads – a situation she encountered recently, she says, at a major British high street bank.

Still, Ingres is working to improve its performance credentials. In July 2009, the company announced that it is working with a Dutch research company called VectorWise on technology that exploits little-used features in microprocessor chips to speed up data-handling processes by a factor of ten.

“Historically, massive advances in chip technology have not always led to massive advances in application performance,” says McGrattan. “This project aims to address that.”

The first Ingres products based on this work will be available in 2010, she says. Other open source databases refute that their products aren’t suitable for mission critical deployment.

Even if that were true, it would be no bar to the cost-saving opportunity that they represent to a broad swathe of organisations and industries, says Ed Boyajian, CEO of EnterpriseDB, which supplies a commercial version of the PostgreSQL open source database, called Postgres Plus.

This is specifically engineered as a compatible, but open source, replacement for Oracle databases.

“By choosing to use a proprietary database as the de facto standard,” he says, “many organisations are paying for numerous instances of the most expensive technology available across the board, even when less than half of their enterprise systems might be truly described as mission-critical.”

Other barriers to adoption include the limited availability of packaged applications that run on open source databases and third party tools to manage them.

Application giant SAP, for example, has no plans to enable its software to run on open source databases, according to Jacob Klein, the company’s vice president of solutions management for data and analytic engines. In October 2007, it took its own, home-grown open source database, MaxDB, back to a closed-source model after seven years ‘in the wild’, although it is still offered free of charge to customers.

But plenty of smaller, independent software vendors (ISVs) think differently on the issue, according to Boyajian of EnterpriseDB. Around one third of the company’s annual revenues (undisclosed by the company) come from ISVs that are porting their applications to PostgreSQL or bundling the database with their applications as part of a broader ‘solution stack’.

“They’re doing this because customers, especially small and medium-sized enterprises, are asking them to,” he says.

Competition intensifies

The commercial opportunity at stake, it seems, is huge – and the attacks and counterattacks between competing vendors continue to come thick and fast.

In April, IBM partnered with EnterpriseDB to embed EnterpriseDB’s Postgres Plus into IBM’s DB2 9.7 database, a move that enables Oracle customers to migrate to DB2 far more freely.

In May, Ingres partnered with open source operating system (OS) company RedHat to offer the Ingres Development Stack for JBoss, a pre-integrated package comprising the Ingres database, Red Hat’s Enterprise Linux OS and its JBoss open source middleware product.

Some industry watchers have speculated that a merger between the two is a distinct possibility. And in August, EnterpriseDB announced an Oracle Migration Assessment Program, designed to “aid the thousands of companies locked into the Oracle database and related products who face a constant drumbeat of price increases from the vendor”, according to Boyajian.

That move came after Oracle bumped up the prices for some components by more than 40%, as detailed in the July 2009 edition of its price list. Clearly, the market is hotting up.

“In some ways, it feels very much like the days of the database wars all over again. You’ve got lots of players, all jockeying for position and competing ferociously on business models and pricing,” says McGratten.

But there is one principle difference, she jokes: “This time around, we’re going to make sure Oracle doesn’t win!” 

Monty Widenius is understandably frustrated, but determined to fight on. As the main author of the original version of the open source MySQL database, he spent 15 years at the helm of MySQL, before leaving shortly after its January 2009 takeover by Sun Microsystems.

Since then, MySQL has changed hands again, with the announcement that database company Oracle is to buy Sun – and Widenius knows this could be bad news for his brainchild.

“Oracle has been silent about what they plan to do. I have asked them both privately and publicly, but received no answers at all,” he says. “Until Oracle shows its card, the only feeling one can have is uncertainty for the future. This is a feeling that seems to be shared by customers, users and MySQL developers alike.”

Still, Widenius has been busy since his departure, launching a new company Monty Program Ab, which has used the MySQL code base to build the MariaDB storage engine.

Monty Program now employs 20 people and is actively hiring more, he says.

The first general availability release of MariaDB is due out in September and is being put together by many of the original MySQL core developers.

“My vision of Monty Program Ab is that it will be the centre of MariaDB development, done in a true open source spirit,” says Widenius. In the meantime, the MySQL community has started to fragment rapidly since the news of the Oracle acquisition broke.

As well as MariaDB, another intiative, Drizzle, led by former MySQL employee Patrick Galbraith, is aiming to shed some of the ‘feature bloat’ that its developers believe crept into more recent versions of MySQL, to produce a more lightweight database server aimed at cloud computing and web applications.

As a result, many developers are now asking which branch of MySQL can be considered ‘official’ these days. Some are even questioning whether they should stick with MySQL in any of its forms, says Ed Boyajian, CEO of EnterpriseDB.

“We’ve seen a sharp increase in interest in migrating from MySQL to PostgreSQL – it’s the subject of the most downloaded white paper on our website,” he says. 

Related Topics