InfoQ interviewed Nik Wekwerth of db4objects (db4o), about his view on the state of ODBMS and new features in the recent db4o release. The discussion started with the enumeration of some of the drivers behind the seemingly increasing popularity of object databases within the developer community. In Nik’s view ODBMS are back and their adoption is growing fast due to 3 factors:
- ease of native access in Java and .NET
- suitability for specific problem domains
- open source
...Developers don't have to learn a new language, [in the db40 case] they just drop in a .jar or .dll file into their project and are up and running... Simplicity of use and full integration to the programming language, the native approach, are key to the success of object databases in general...Queries that are native to Java or .NET leverage benefits of development tools such as type-safe coding, eliminate the mismatches that surface when complex structures are used with object-relational mappers, and save the effort of learning the intricacies of another language.
...The requirements of the application should define whether a developer should use an rdbms or an odbms. Many application domains are better suitable for object oriented structures and persistence, such as complex configuration or security data, searchable multi-media or spatial data, and very complex relationships inherent to a data domain. Another determinant is whether persistence is fully “embedded” into the application, i.e., invisible to the end user, in which case an object database is much cleaner and easier to maintain than an application that has to map objects to tables.
...open source is a big driver in the recent surge of adoption of object databases. Developers have actually always liked the idea of an ODBMS. It was more the DBAs and managers that were bought into RDBMSs like Oracle, be it for the right (data abstraction) or the wrong reasons (golf tournaments). Open source has put developers back into a more powerful position. They can now more autonomously make design decisions “below the radar of management” and prove that object databases helps them save time, build more compelling architectures and ultimately save their customers a lot of money.
db40 has seen a lot of adoption for scenarios where it is embedded within an application, with no separate deployable database visible to end users or developers. Nik characterized these as zero-admin environments where an object database' simplified API, elimination of any run-time administration, and open source properties make it very suitable for embedding into not only devices like robots, cars, and cell-phones, but also packaged software, SCADA systems and caches anywhere in the network. Nik mentioned that Ricoh is using them in their printers & copiers.
Commenting on adoption in the enterperprise, Nik cited an example of ow financial institutions leverage db4o as an intelligent cache of data from their central database to enable faster transactions. He also added that object
databases can live nicely side by side with rational databases because, as in db4o’s case, they allow data to be replicated into the relational database.
The discussion shifted afterwards to the new features in db4o v7.0:
The biggest highlight is Transparent Activation (TA). TA simplifies object retrieval by automating object activation. With TA, enabled objects are only activated when they are requested. The key here is that developers don't have to define the activation depth manually any more, which simplifies coding. Another benefit is that memory consumption is optimized, as objects are only loaded into memory when they are requested.
With Transparent Activation developers no longer need to worry, when and how objects are loaded. db4o has more information at runtime than developers have at development time, so it can do a smart job and only load the objects that are really needed.
The commercial-only offering of the product includes 2 other important features:
- ObjectManager Enterprise - a plug-in for Visual Studio which enables database browsing and querying from within the IDE.
- XtremeConnect - a virtual pairing technique that allows for remote pair programming via Skype and TightVNC.
db4o is also preparing for an upcoming release of the product for the .NET platform. He described the similarities and differences between the Java and the .NET versions:
Both our Java and our .NET version are produced from a common shared code base. All our distributions are produced in a fully automated continuous build that converts code between platforms. In some places we do make some adjustments for the respective platforms, because we follow standard conventions...Only a few differences apply: The current db4o Replication System (dRS) is build on Hibernate and is more complete for Java than .NET. On the .NET side we are in the process of providing a LINQ interface shortly.
As a result, Nik mentioned that “the database file format is the same. You can open databases with .NET that were produced with Java and vice versa. You can also connect to a .NET server from a Java client and vice versa.”
The interview concluded with Nik sharing his vision in regards to the opportunities that Google’s Android platform brings to their product. db4o allows ISVs that build software for Android to develop:
...client-side applications that can adapt automatically over time (as db4o detects and adjust to schema changes) and can collaborate and share data among different applications. db4o can take care of all aspects of data administration, e.g. when you update your software with a new data model over the air.
db4o also works directly with members of the Open Handset Alliance to overcome the database shortcomings of Android and provide powerful cross-application persistence enhancements. As a result, for instance, mobile carriers can provide exciting new services around user generated data (UGD) such as complete backup functions of all application data. If you loose your handset, you can get a new handset and all your personal data is already preloaded. Another use case is p2p data sharing and p2p social networking.
What is your opinion? Are you using an ODBMS in your applications? What are the scenarios in which you think an odbms works better than the traditional rdbms approach?