Business Factors in OSS Database Companies

After many years of slow and steady growth, open source software (OSS) has begun to make an impact on corporate culture. Online forums abound with discussions about how open source projects are changing the way businesses use software. Venture capital companies now look for open source projects to invest in and larger IT companies like IBM and Sun Microsystems are turning to open source to gain a competitive edge. OSS is a fast-growing influence and yet the topic of how business affects OSS companies has been rarely studied.

Executive Summary

This report provides one of the few original studies on how business influences affect the development of companies that choose an open source business model. It is an objective attempt to compare, measure, and contrast business issues particular to OSS database companies. By determining the core influences of companies doing business completely in the OSS database realm it should be possible to isolate these factors and understand how they apply to other open source companies.

This study is based on original research conducted in cooperation with the principles of established OSS database companies. Six OSS database companies participated in the study: MySQL AB, IBPhoenix, Sleepycat, DB4Objects, Great Bridge, and Genezzo. They were selected for study specifically because they focused solely on selling and supporting open source databases.

Our study methodology consisted of formulating a variety of questions that would characterize aspects of a company’s business model in terms of revenue sources, size and growth, business emphasis, customer satisfaction, suitability of the product, and type of licensing. The specific results to these questions are available in a separate report.

I studied four active commercial OSS database companies, MySQL, IBPhoenix, Sleepycat, and DB4Objects. I also included in the study one important failed commercial venture, Great Bridge, and one pre-commercial venture, Genezzo. Although I considered the PostgreSQL project, the project was not included because it is not supported by any one specific company. Backplane was invited to participate but declined.

Of the business influences studied, the biggest influences were:

  • Whether the license was open or dual-use.
  • Time along the open source adoption curve.
  • Growth capacity.

    These influences are discussed in further detail in the results and analysis section at the end of this article and discussion is provided of how these business factors extrapolate to general open source companies.

    Company overview: MySQL

    MySQL AB was formed in 1987, and by Marten Mickos joined the company in 2001. MySQL employs 220 people worldwide. MySQL had $20M in gross revenue for the last fiscal year and 100% growth for the past 3 years running. The company has grown from 60 to 220 people in the last two years. The company estimates it has 6 million installations worldwide with another 40,000 downloads each day.

    In June of 2003 MySQL secured $19.5M in Series B financing from Silicon Valley venture capital firm Benchmark Capital. Benchmark was chosen as a funding partner because of its track record with other companies like eBay and Red Hat. When asked why he decided to take the path of venture capital funding, Mickos replied, “Because we have the potential, because we can grow to venture capital expectations.”

    When asked if MySQL was on the IPO track Urlocker responded, “That is the preferred track. We would want to see $50 – $100M in annual revenue before considering an IPO.”
    65% of MySQL’s revenues come solely from licensing the database. 35% of revenue comes from training, support, and documentation. These percentages are likely to change as the company moves to a more scalable support model.

    In a commodity market, service is a key differentiator. MySQL has done well in support and can offer 30 minute support turn-around. Urlocker cites a recent gold member support incident which resulted in a fresh, tested database patch within 24 hours. “Our developers are key to our support. With larger database companies your primary tech support contact knows less than you do.” Even with this edge the company will continue to invest in the area of support.

    MySQL’s business emphasis is to help companies lower their TCO for their database investment by specifically providing a low cost commodity database. MySQL wants to deliver a database solution an order of magnitude cheaper than closed source databases. Says Zack Urlocker, VP of Marketing, “Databases are well understood so we don’t have to spend time explaining the product.”

    MySQL’s biggest challenge

    The company’s biggest challenge is managing growth. High growth makes scalability a real challenge and creates the possibility of exceeding management’s capability. Another challenge is meeting the needs of unsophisticated corporate users. “Our product has to meet the ’15 minute rule’,” says Urlocker on the company’s support challenge. “If the user doesn’t have MySQL installed and running in 15 minutes we haven’t done our job. Our support must be top notch. A year from now our challenges will be different but the small issues we encounter can present special challenges during high growth.”

    “MySQL uses a ‘dual licensing’ business model. Under this model, users may choose to use MySQL products under the free software/open source GNU General Public License (commonly known as the “GPL”) or under a commercial license.

    “With the GPL license, MySQL is available free of charge. Users may download the software for free and modify, integrate and distribute it. However, GPL users must abide by the rules of the GPL, which stipulate that if a MySQL-based application is redistributed, the complete source code for this application must also be open and available for redistribution.

    MySQL AB offers a commercial license for organizations that do not want to release the source code for their application. Commercial license customers get a commercially supported product with a level of assurance from MySQL AB, without the requirement that their MySQL-based software must be ‘open sourced’. MySQL’s dual licensing increases freedom in two ways: first, it encourages the growth of free software by licensing MySQL under the GPL; second, it makes it possible to use our software in situations where the GPL is not applicable.”

    The dual licensing model is a powerful vehicle for enhancing revenue. Without the ability to offer a commercial license for the core database product, MySQL would forego fully 66% of its total revenue. In essence, the dual license allows MySQL to triple its revenues.

    Company overview: IBPhoenix

    IBPhoenix employs 6 people, grossed $50k its first year, and has nearly doubled revenue each year since. Beach formed InterBase Software to commercialize the existing and mature code base that Borland had taken open source after several years of closed source development. Borland continues to sell InterBase as a closed source product. InterBase is a relatively small company but it is profitable, it has a solid international presence, and it is growing rapidly.

    The InterBase / Firebird database project has a colorful history. The initial InterBase product was started in 1984 by Jim Starkey, a former database architect for DEC. IBPhoenix does business as InterBase Software Ltd. and was founded in 2000 by Paul Beach, its president and managing director.

    IBPhoenix licensing

    The InterBase database product was sold as a closed source product by three companies before Borland took it open source. When Borland subsequently cancelled the open source project, two of the three managers formed the existing InterBase software company to maintain the open source product. The InterBase project is now named Firebird after the mythical phoenix that is re-born anew from the ashes of its funeral pyre. But the new open source project is not without serious legal entanglements.

    IBPhoenix uses an open and unrestricted license for the InterBase product. The core InterBase source code is owned and licensed by Borland under the terms of its InterBase Public License (IPL). Borland’s IPL license is similar to the GPL in that any code developed by a commercial interest must be made available to the public at no cost. IBPhoenix complies with the IPL by making all of its source code open without any restriction.

    Because of this unique ownership situation IBPhoenix does not have the option of offering the InterBase database under a commercial license. It can only sell support, contracting, documentation, and add-on products.

    Company overview: Sleepycat

    Sleepycat was formed in 1996, but has it roots in the initial release of Berkeley DB in 1991. Michael Olson is the president and the company currently has people 25 on staff and out-sources several additional positions. The company’s gross revenue has grown 30% every year since inception, with a 60% increase last year. In 2004 the company saw about 50% growth in profits. By all accounts the company’s Berkeley DB product is a highly successful product in a well-guarded product niche.

    Fully 65% of the company’s revenues come from licensing its Berkeley DB product. Roughly 30% of revenue comes from selling releases with an annual support contract. All engineers are available to work on open customer support issues. Selling services is not a focus of the business. Other avenues of revenue for Sleepycat include selling complementary products for XML and Java support.

    Sleepycat’s biggest success

    Sleepycat’s biggest success is the ubiquity of the Berkeley DB. The company claims an estimated 200 million deployments. Every copy of Linux has a version of Berkeley DB in it. This success is directly attributed to the company’s dual license strategy and the product being open source. The widespread distribution of Berkeley DB creates a large market to up-sell, and takes sales away from competitors. This is a classic example of the benefits of the open source distribution model. The free availability of its source code creates a serious barrier to entry for any competitors.

    Like MySQL, Sleepycat uses a restricted, dual-use license for Berkeley DB. The terms governing the use and distribution of it are briefly described here: “Your application must be internal to your site, or your application must be freely redistributable in source form, or you must get a license from us.”

    Sleepycat does not have to worry about competing for sales or service against companies using its source base because this is prohibited by the license. Sleepycat negotiates every single license, and will only engage in licensing deals that prohibit competitive licensing.

    Like MySQL, Sleepycat’s dual-use licensing allows it to triple its revenue potential by commercially licensing its database product. Selling services is not a business focus for Sleepycat.

    Company overview: DB4Objects

    DB4Objects was incorporated in 2004 but has a software history going back to 2000 when Karl Rosenberger began the coding. Christof Wittig is the President and the company has a worldwide staff of 20 people. The company’s gross revenue in its first year of business is roughly $1M and has 10,000 members in its registered user community. The company is funded in part by angel investors that Wittig met while attending the Stanford Graduate School of Business.

    Revenue emphasis is primarily on licensing the database product which accounts for 80% of the revenue. Annual support contracts account for the 20% balance. The true focus of the company is on product sales and the OEM business service is the only means of achieving licensing.

    DB4Objects biggest challenge

    Wittig says that his biggest challenge is to educate the industry that object oriented databases are not dead or lagging and that DB4Objects offers an entirely new angle for providing zero maintenance embeddable database solutions. In the past, commercial offerings of object oriented databases such as Versant and ObjectStore offered performance but were expensive and complicated to manage. Object relational mapping solutions for relational databases such as Toplink and Hibernate bridged the object oriented languages with relational databases but incurred a performance penalty for Java and C# users.

    DB4Objects touts its top position in the PolePosition open source database benchmark as evidence of its suitability in high performance embedded solutions.

    Wittig says that although DB4Objects competes with SleepyCat and IBM’s Cloudscape, his biggest competitors are developers who will often write their own database persistence. Wittig sites as competitive advantages the ability to provide a “One-Line-of-Code-Database” which provides substantial performance over relational database performance. Its 400k footprint and automatic schema versioning are also advantages for embedded solutions such as cell phones and PDAs.

    DB4Objects uses the standard dual-use GPL license for its product. It recommends the commercial license when developers compile commercial software using db4o, if non-GPL’ed software contains specific references to the db4o software, or if non-GPL’ed software requires db4o to work.

    Company overview: Great Bridge

    The story of Great Bridge is a fascinating tale of business failure despite technical excellence, ample funding, and previous business success.

    Great Bridge was formed in 1999 and closed in 2001. The company had roughly 45 people on staff at its peak, and spent $16M during its 18-month existence. It was formed by Frank Batten, Jr., CEO of Landmark Communications. Batten was an early angel investor to Red Hat and hoped to duplicate his open source business success with an open source database.

    The initial business idea was to hire the original PostgreSQL developers to take the mature PostgreSQL code base commercial. Although a great idea in theory, it was difficult in practice. Differences of opinion on the direction of the company surfaced early and the first CEO left after only two months on the job.

    The original PostgreSQL core developers were paid to do full time what they had done previously as part time developers. One developer fixed bugs and did core development. Another developer became project spokesman and arranged public events. The third developer worked on new features.

    Great Bridge set out to create a business around the existing and mature PostgreSQL code base and to do what Red Hat had done with Linux. PostgreSQL had traditionally been available as a free product and the company saw little value in charging licensing fees for what people could download elsewhere for free. The company’s business emphasis changed frequently in an effort to find a profitable niche. The frequently changing business focus is cited as the company’s single biggest problem. The changing emphasis, however, could have only been a symptom of an unprofitable company.

    The company initially offered tiered support options. The company also offered hourly and 24×7 support options, but support did not sell well. It provided some consulting for porting & performance tuning. Aside from the selling, licensing, and servicing the basic database, the company did not seek any other avenues of revenue.

    According to co-founder Ned Lilly, “It’s definitely the case that Great Bridge, like every other company that got money at that time, could have spent it better. Generically, the idea was to throw a lot of money into a company, grow it fast, take advantage of the market, and then cash out. To some degree Great Bridge got caught in the dot com fever of the time.”

    Lessons Learned from Great Bridge

    The history of Great Bridge provides valuable lessons for open source entrepreneurs with a big idea. Investors come as a package deal. They have specific business models, plans of execution, and exit strategies. When you are looking at taking investment, it’s important to look at how investors will be involved in the management decisions, whether directly or by appointing personnel to your company.

    Another valuable lesson is that an infusion of venture capital funding and a large staff will by no means guarantee sales of a product that has insufficient demand.

    Company overview: Genezzo Systems

    Genezzo Systems was formed in 2002 by Jeffrey Cohen. He is the president, and the only full time employee. The company currently exists in a pre-commercial state and has not yet licensed or sold any products. Genezzo Systems has developed the prototype of a next-generation distributed database architecture. The Genezzo flexible architecture was inspired by Cohen’s frustrations as a developer at Oracle.

    Virtually all aspects of Genezzo are designed for pluggable extensions, and many are capable of runtime changes. This enables architectural designs for massively parallel database capable of scaling beyond current commercial options. It also allows easy addition of scalable features such as encrypted tables, inexpensive replication and failover.

    The project is in the early phases of organization and has yet to commercialize its database product. The motivation for including Genezzo in this study was to capture the pre-commercial attitudes and perspectives that many developers have before going into a commercial venture. This is important because many technology projects start as a pre-commercial or non-commercial venture and then graft on a business model after the fact. The roots of such tech startups can have important effects on its future business ramifications.

    Genezzo’s anticipated business emphasis is to provide a high end extensible and customizable distributed database system. It is targeting the fields of bio-informatics, data cleansing, and textual analysis. These application domains present problems for existing mature relational databases and thus an opportunity and niche for Genezzo’s next-generation database features.

    Cohen anticipates providing professional services for custom functionality development. These features would include customized index methods. The company would look at ways to develop specialized data mining functions. An example would be special matching functions for genetic analysis of genome databases.

    Genezzo also anticipates starting with a variety of support packages including 24×7 highest tier support package, pay per call support, and prepaid long term support contracts. In addition to selling, licensing, and servicing the basic database, the company will seek revenue from training and documentation.

    At the beginning of this study Genezzo had not yet chosen a licensing scheme for the software, but was leaning towards an open and unrestricted license. Based on the results of this study, Genezzo made the decision to adopt a dual-license scheme with the GNU GPL.

    Impact of the Licensing Model

    This study found that the licensing model had a major impact on a company’s core revenue source. The dual licensing of MySQL, Sleepycat, and DB3Objects enables them to triple their revenue by selling a commercial version of their product. This is the basis for a large part of the respective companies’ product revenue.

    There are financial consequences to not adopting a dual license. When Great Bridge adopted the existing PostgreSQL product it also adopted the existing unrestricted BSD license. This made it difficult for Great Bridge to sell a product that was otherwise available free and with no license restrictions. Its revenues came mostly from support. The same was true for IBPhoenix which adopted the existing InterBase product but was prohibited by Borland’s IPL license from licensing its own product.

    Not having a dual license puts an open source company at a great financial disadvantage. Adopting the unrestricted BSD license means foregoing fully two-thirds of possible product revenue. Not having this revenue means a company must be very careful managing its cash flow and that it cannot sustain rapid growth.

    The open source adoption curve

    The successful companies each followed a similar growth curve dictated by its open source adoption. This adoption curve has unique characteristics that distinguish open source products from the classic product introduction curve.

    The open source adoption curve is driven by factors such as community involvement, word of mouth, forum discussions, ideological concerns, cost, as well as suitability to task and product alternatives. A successful open source adoption curve is characterized by a long, slow beginning as awareness of the product spreads by word of mouth to interested parties, using what economists call the “network effect.” After a critical users mass is achieved the product and company sees a much more rapid growth in a shorter timeframe.

    A classic product adoption curve is driven by factors such as advertising, street promotion, distribution through traditional supply chains, manufacturing capacity, and existing products of similar function. A classic product adoption curve can be accelerated simply by spending money on advertising and marketing activities. Open source companies typically do not have large advertising budgets to accomplish the same means of promotion.

    SleepyCat and MySQL each experienced many years of slow but steady growth before seeing rapid growth. The InterBase/Firebird database product had a curious hybrid adoption curve since the product started as close-source, was open sourced by Borland, which then stopped releasing updates under the IPL, and development under the IPL was taken up by IBPhoenix. Its growth is characterized as steady and moderate.

    Although all open source companies like to brag about their download numbers as evidence of adoption, the correlation between downloads and actual sales is hard to establish. When a company boasts of a large download number for their product they can really only brag about the general interest in their product. After an open source product is downloaded, the end user must still make the decision to license the product.

    Influence of growth capacity

    The circumstances of a company’s formation can affect its growth capacity and set the stage for later success or failure. Three success stories include Sleepycat, MySQL, and IBPhoenix contrasting Great Bridge as a failed venture. DB4Objects shows the initial signs of success but has not yet posted a profitable year. The successful ventures each followed a similar course; each company started as a small but profitable company that continued on a slow developing growth curve and then experienced rapid growth.

    The slow growth period provides time to build support networks while product awareness spreads. This in turn provides growth capacity which contributes to the subsequent rapid growth.

    Great Bridge started with a pre-existing code base and the infusion of $16M in investment money. By Batten’s account their strategy was to grow the company first and seek profitability second. This is a typical high stakes venture capital strategy. Said Batten about the experience, “We assumed the business was going to be successful and staffed accordingly. In retrospect we should have started with 5 or 6 people, gone to profitability, and then grown.”

    Two issues of Great Bridge’s formation set it apart from the three successful ventures. It started with a product it could not license, and this decreased its potential revenue. It grew so quickly with investment capital that it could not reasonably cover its costs and achieve profitability in its short lifespan. Great Bridge exceeded its growth capacity, was unable to accelerate its growth curve, and ran out of operating capital.

    IBPhoenix is well into its growth curve having achieved a greater product awareness and a large customer base, and benefits from a mature support network. Although its profitability is handicapped by its inability to license the database product, IBPhoenix has been careful to manage its cash flow and continues to grow at a moderate rate.

    Conclusions

    There are some important lessons here for companies looking to adopt an open source business model or invest in an open source company. First and foremost, choose a dual or GPL license so you can charge for the licensing of the product and not just charge for support services. Choosing an open, or BSD license means foregoing 66% of your possible revenue.

    It is important to understand the factors influencing the open source adoption curve. This curve traditionally extends over several years as product awareness spreads slowly. It should be possible to accelerate the adoption curve with traditional marketing methods, but the long duration required could be troublesome for companies partnering with investors hoping for short payback periods. This long curve will undoubtedly prove fatal for some of the new generation of OSS startups that have partnered with traditional venture capitalists working with template investment models.

    As with all companies it is vital to understand the true growth capacity for a product and to understand its suitability to task. All entrepreneurs should have an unshakeable confidence in their products, but not having thought through the business model or understanding the growth capacity will most definitely prove fatal to a new OSS venture. Running an OSS company means mastering a very lean budget and having saint-like patience.

    About the author:
    Andrew Hudson is a freelance IT consultant who occasionally does research on database companies.


    If you would like to see your thoughts or experiences with technology published, please consider writing an article for OSNews.

  • 10 Comments

    1. 2006-02-28 1:24 pm
      • 2006-02-28 3:06 pm
    2. 2006-02-28 1:28 pm
      • 2006-02-28 1:54 pm
    3. 2006-02-28 5:33 pm
    4. 2006-03-01 4:05 pm
    5. 2006-03-01 5:58 pm
    6. 2006-03-02 1:27 am
    7. 2006-03-03 3:42 am
    8. 2006-03-03 8:27 am