OWLIM comes in three editions: OWLIM-Lite, OWLIM-SE (Standard Edition) and OWLIM-Enterprise. The editions share the same inference mechanisms and semantics (rule-compiler, etc). OWLIM-Lite is a fast, in-memory repository, whereas OWLIM-SE uses a scalable, file-based indexing mechanism. OWLIM-Enterprise is a clustered infrastructure based on OWLIM-SE.

OWLIM-Lite (previously SwiftOWLIM) is designed for medium data volumes (below 100 million statements) and for prototyping. Its key characteristics are as follows:

• reasoning and query evaluation are performed entirely in main memory
• it employs a persistence strategy that ensures data preservation and consistency
• the loading of data, including reasoning, is extremely fast
• easy configuration.

OWLIM-SE (previously BigOWLIM) is suitable for handling massive volumes of data and very intensive querying activities. It is designed as a commercial-grade database management system. This has been made possible through:

• file-based indices, which enable it to scale to billions of statements even on desktop machines
• special-purpose index and query optimization techniques, ensuring fast query evaluation against very large volumes of data
• optimized handling of owl:sameAs (identifier equality) to boost efficiency for data integration tasks
• efficient invalidation of inferred statements, which allows for efficient delete operations

OWLIM-Enterprise (previously BigOWLIM Replication Cluster) is designed for resilience and parallel query-answering performance through:

• parallel execution of queries on multiple worker nodes
• dynamic configuration of cluster
• automatic fail-over and synchronization - no single-point of failure

The different editions of OWLIM are designed and tailored to efficiently meet the wide range of requirements for RDF databases and reasoning engines.

OWLIM Versions

The following table summarizes the differences between the versions that are currently supported.

Supported Features

There follows some brief explanations of the features supported by the various editions of OWLIM. More information is available in the relevant sections of the User Guides for the corresponding editions.

Named graphs (NG) represent an extension of the RDF data-model, where quadruples <s,p,o,ng> are used to define RDF multi-graphs. This mechanism allows for (among other things) handling provenance when multiple RDF graphs are integrated into a single repository.

Sesame 2.x is the latest official release of the most efficient RDF framework that supports named graphs (under the name contexts) and SPARQL. The versions of OWLIM that support SPARQL are compliant with Sesame version 2.x. More information about the APIs for using OWLIM is available on the Access Methods page.

Jena is a popular framework for RDF management that includes the ARQ SPARQL engine. This engine already supports a number of proposed extensions for SPARQL (aka SPARQL 1.1) including aggregations and update instructions. More information about the APIs for using OWLIM is available on the Access Methods page.

SPARQL is the most popular RDF query language (W3C recommendation). Comprehensive support for SPARQL requires named graph support.

The owl:sameAs optimization delivers dramatic improvements in performance and usability when huge volumes of data from multiple sources are integrated.

Full-Text Search is a powerful technique that allows 'stringized' resources and literals to be searched using either a proprietary term search technique or using the integrated Lucene text indexing and search engine. Full-Text Search is integrated with SPARQL to allow powerful hybrid queries.

Geo-spatial extensions use special query constructs and SPARQL extension functions to allow fast indexing of entities with 2-dimensional co-ordinates on the surface of a sphere (the Earth). Special indices allow queries to quickly find points within rectangles, polygons and circles and to compute the distance between points.

RDF rank, similar to Google's PageRank, can be calculated for the nodes in an RDF graph and used for ordering query results by importance, visualization and many other purposes.

Efficient retraction of statements: OWLIM-SE stores explicit and implicit statements; the latter are inferred from the explicit statements during loading (an approach known as materialization). When explicit statements are removed from the repository, any implicit statements that rely on a removed statement (i.e. which cannot be inferred in its absence) must also be removed. With OWLIM-SE the removal of explicit statements triggers a unique algorithm that invalidates only those inferred statements that can no longer be derived in any other way. The time required for invalidation after deletion of a group of statements is comparable to the time required for inference when the same group of statements is inserted into the repository - this massively improves the efficiency of statement deletion when compared to the typical behavior of other engines using materialization that invalidate all the inferred statements and then re-compute them all.

RDF priming is a technique based on activation spreading that allows efficient data selection and context-aware query answering when handling huge datasets.

Notifications provide a mechanism for clients to react to statements in the update stream, i.e. to be notified when statements are inserted in to the repository that match previously requested statement patterns.

Replication Cluster (RC) is the major component of OWLIM-Enterprise that allows OWLIM nodes based on OWLIM-SE instances to be used in parallel to increase query answering performance and improve resilience. Master nodes perform load-balancing and synchronization functions, as well as allowing dynamic configuration of the cluster. An RC configuration can support further advanced features, such as online backup, and higher levels of transaction isolation, namely Serialized. It is presented in greater detail here.