The Free Database Niche
The following article appeared in Computerwoche, a German weekly IT newspaper a few weeks ago. The article that was printed was originally in German, this is an English translation of the original article. Despite emails and even a letter responding to this article we have had no response from Computerwoche or Wolfgang Sommergut. So we decided to print the article on our web site along with our response so supporters of Open Source databases, and in particular Firebird can see what we had to say in our defence. You can read our response here.
The Linux open source success story is a model for free databases. However their main area of application appears initially to be limited to simple web applications.
By CW journalist Wolfgang Sommergut
One of the basic assumptions of open source theorists has long been that free software will, more than anything else, revolutionise IT infrastructure. In particular, it is supposed to radically alter business with operating systems, web or mail servers as well as databases. Whilst Linux and Apache seem to confirm this prognosis in the first two categories, open source has obviously only minor effects on the database market. At least the large producers do not believe the free competition capable of gaining a significant share.
Producers donate code
At first glance, this attitude may seem incomprehensible, as the open source warehouse has made considerable progress this year. IBM and Computer Associates were two large producers to release their product code: Big Blue donated Cloudscape to the Apache Software Foundation, CA declared Open Ingres as open source software. Furthermore the open source veteran, PostgreSQL is on the verge of releasing its version 8 which, in addition to the previously supported Unix systems, also incorporates a native Windows version. Moreover the Firebird team published version 1.5 of its database, and MySQL 4.1 is about to be released.
The big three in the database business, Oracle, IBM and Microsoft reacted some time ago to the challenge of the free competition. They secured their business downwards with inexpensive light versions of their products. Microsoft has even provided a gratis SQL Engine with its Microsoft Database Engine (MSDE), which developers can integrate into their applications and distribute it with their software without any restrictions. The company has announced an Express Version for the SQL Server 2005, which will be considerably upvalued in comparison to the MSDE. It is to have a graphical front-end, studded with wizards, and will also profit from the innovations in the full version. The SQL Server Express comprises complete XML support including Xpath and Xquery, allows programming of stored procedures in all .NET languages and can be administrated using the Microsoft Operations Manager.
IBM has taken a similar route with its Express version of the database flagship DB2. It is subject, as are its competitors' starter versions, to certain hardware limitations such as the number of CPUs or the usable memory. Furthermore several high-end features are not included, these being reserved for the large versions. However Big Blue's product range is, due to its heterogenic product portfolio, somewhat inconsistent. The Express version is only available for Linux and Windows, and for none of the IBM systems. If users outgrow the small version, they may have to face a change of platform. This also applies, by the way, for the free version of Adaptive Server Enterprise from Sybase, which is only available for Linux.
The Edition One from Oracle is more concise. This is also a light version of the 10g database with hardware limitations. If a higher performance is required, the Standard and Enterprise editions are available for the same operating systems. If the continuous success story of Linux or Apache is taken into consideration, then the question arises, why light versions, of all things, should deflate the open source competition. The web server example illustrates that the free full version of Microsoft's Internet Information Server was not able to stop Apache's triumphal march. In comparison, the light versions of the large databases seem less attractive. IBM and Oracle charge license fees, Microsoft's offerings require the purchase of a Windows server. Nevertheless there are still some arguments that speak for the large suppliers.
Representatives of all three companies name the lack of, or poor development of the free databases as a fundamental reason. The leading producers' light versions may cut back in the range of features, but their core is designed for enterprise deployment and they have established themselves in the long term. They offer modern administration tools and can be incorporated, via the necessary agents, into large system management frameworks. The big three, as infrastructure companies, enable database administration using the same console, from which the majority of other components for the respective platforms can also be administrated. In contrast, the majority of free databases are limited partly to simple web front-ends. Furthermore, all light versions with functionalities such as auto-tuning, benefit smaller companies without their own database administrator (DBA).
If required, users can transfer their software to the larger versions, which offer all those enterprise features that most of the free databases don't, without having to make any adjustments. These include, for example, advanced cluster functions, standby databases, Java or .NET support for stored procedures or integrated storage management. In spite of the abstract possibilities that standards offer for data access (JDBC, ADO, ODBC, etc.), changing DBMS demands a considerable amount of time and effort. Incompatible SQL dialects or differences in datatype are responsible for this.
The big producers do not only offer scope for advancement or upgrading in the traditional areas of application. For some time now, a trend can be observed moving from mere relational databases towards platforms for data management. This can be seen in the integration of OLAP, data mining or reporting. Alongside this, the leading products are continually becoming better suited as storage for documents of all kinds. This category also includes extensive XML features.
Even here it could be argued, by taking a look at Linux, that a large open source community can catch up on the shortfall compared to commercial projects. Linux however is lucky to have the energetic support of numerous companies, including IBM. This constellation does not exist with databases. The nearest to an exception is the Swedish MySQL, that received a technology transfer in the form of Max DB from SAP. The company from Walldorf has a certain interest in the existence of a free database alternative for their own ERP solution. After all, Oracle and Microsoft, as two of the large database producers, are also strong competitors in the field of managerial software.
Know-how for free teams
Without the backing of large companies, the required database knowledge, necessary for free teams to be able to build up a veritable competition, can hardly be organised. This is apparently even scarcer than the know-how necessary to develop a compiler or a graphical user interface. This can be illustrated by the fact that those people who originally launched Interbase, and finally landed at Borland via the acquisition by Ashton Tate, are still at the helm of Firebird.
This deficiency is even more explicitly manifested at MySQL, which is practically solely developed by the Swedish company's fixed programming team, and therefore does not follow the development model of open source. With limited resources and without recourse to a large community, it is difficult to stand up to Oracle, IBM or Microsoft. So the update cycles of MySQL are at least no shorter that those of Oracle, and that taking into consideration the current stage of development where standard features, such as triggers, views or stored procedures, are still absent. In view of the limited capacities, MySQL follows an acquisition strategy including, for example, the purchase of the InnoDB Engine, or the cluster solution acquired from Alzato. Together with the parallel development of MaxDB the organisation has accrued a considerable integration challenge.
Questionable cost advantages
Although Oracle and IBM charge licence fees for their starter versions, they are not necessarily more expensive than the open source competitors. MySQL is subject to the General Public License (GPL), but only when the database is part of an open application. Otherwise the commercial license applies with its admittedly moderate fees. In conversation with Computerwoche, Kaj Arnö, Vice President MySQL, estimated the share at more than 50% of total volume. However the support costs turn out to be more expensive than the purchase. The best alternative of basic support still costs 12,000 Euros annually. Oracle argues that for a technically superior product, such as the 10g Edition One, this fee is only 22% of the original acquisition cost.
The support situation is similar in the case of other open source or free databases. Sybase charges a minimum of $ 2,200 for ASE Express support. In the case of Firebird as well, the cost for support by the IBPhoenix team rapidly reaches several thousand Euros, and the pricing structures for PostgreSQL lies in the hands of existing service providers.
A further hurdle, when considering deployment in the enterprise, is that companies seldom purchase the mere technology for storing information. Much more often, the choice of database is made at the time of purchasing a specific application. This particularly applies to the lower segment, when small companies require branch-specific solutions or commercial software. These are almost always certified for the databases of the large suppliers, and seldom for their open source competitors.
Integration secures position
The big three support this trend, by saddling a series of solutions themselves onto their data storage. In Oracle's case, this is the Applications or the Collaboration Suite, IBM offers a whole portfolio of Content Management Solutions based on DB2, and Microsoft requires the SQL Server for its Sharepoint Portal Server.
The large companies however do not only integrate their databases with their software infrastructure (as in the case of Administration) and their applications, but also with their development tools. Again Microsoft is notably aggressive in this area, and has announced that a free Express version of numerous programming tools is to be included in the slim version of the SQL Server 2005. Even Oracle's Jdeveloper is closely tuned to its own house database and is free of charge. IBM's Java IDE is open source anyway, and offers plug-ins for the modelling of DB2 databases.
The free adversaries shine however by cooperation with open script languages such as PHP or Perl. They are the natural choice for web applications based on Linux and Apache, i.e. within the LAMP batch. The majority of MySQL's claimed four million users can avail themselves of such configurations. That is the domain of free databases, which is illustrated by the fact that they are part of most web-hosting packages. In this field it will be hard for commercial products to suppress them. However open source databases are likely to remain a fringe entity.
Arguments against free databases:
- The large producers are targeting the lower market segment with inexpensive light versions.
- These "Starter" versions are technically more refined than the majority of free databases.
- Upgrading to commercial enterprise versions requires alterations to the applications.
- The light versions are closely interlocked with the platforms of the large producers (administration, tools).
- High support costs often countermand any savings on license fees.
- Standard software is only seldom certified for free databases.
Free databases form, in combination with Linux and script languages, the basis for many web applications. There are certain hindrances blocking their breakthrough into enterprises. Above all the three leading producers, IBM, Microsoft and Oracle, have secured the lower performance segment with reasonably priced "Starter" versions. The evidence indicates that the success story of Linux will not be repeated so quickly with databases. MySQL & Co. initially remain niche products.