Blog > Post
Multi-model database systems â€“ a new trend!
We have been contacted by several database vendors, complaining that we do not correctly classify their systems. They all were right, their systems are supporting more than one data model.
- MarkLogic serves as Native XML DBMS as well as having Search Engine capabilities
- Virtuoso not only is a relational DBMS, but also can be used as a Native XML DBMS or an RDF store.
- OrientDB and ArangoDB can be seen as graph DBMS as well as document stores.
- FoundationDB (by means of its layered architecture) and FatDB can be used as document stores and key-value stores.
- Sqrrl and GlobalsDB both serve as document stores and as graph DBMS and add wide column store technology and key-value access respectively.
Multi-model database systems indeed seem to be an actual trend in database technology:
We do not like to introduce a new category 'Multi-model DBMS', because apart from being 'multi-model', those systems would not necessarily have anything else in common. However, we react to the trend in allowing more than one data model per system and indicate those systems in our ranking as 'Multi-model'. The information, which specific data models are supported can be displayed by hovering over the small information icon.
Of course, the systems attribute description (e.g. https://db-engines.com/en/system/Virtuoso) shows all supported data models.
With respect to our category classification (which is based on the data model), we consequently include a multi-model system in all categories it’s data models fit into. As a result we see a couple of changes in the category rankings, i.e.:
- Virtuoso takes the lead in the ranking of RDF stores and the second place in Native XML DBMS.
- OrientDB and FoundationDB make it into the top 10 of the document stores ranking.
- MarkLogic becomes number 6 in the search engines category.
Having explained our approach, we would finally like to mention that there are a couple of specifics in it:
- Each RDF store can be seen as a graph DBMS. Therefore we do not explicitly include RDF stores in the graph DBMS category.
- Many relational systems provide some kind of object oriented features (user defined data types, etc.). As long as they do not fully support inheritance, we do not classify them as object oriented dbms.
- Nearly each system could be (mis-)used as a key-value store. We put a system in this category only when it is predominantly used as a key-value store.
Share this page