Very Large Graphs for Jena

Very Large Graphs - v0.1

The very large graph implementation extends the Jena in-memory model, the Jena RDB model, and, the Jena SDB model to give the user an impression that memory is infinite. When we say memory is infinite we mean that we have moved the problem from memory to space to disk space by making the disk space available to the application as virtual memory. To make use of disk as virtual memory we take nodes in the RDF graph and write them to disk in a Lucene index. The nodes to be written to disk are selected based on the common memory management algorithms like FIFO, LIFO, LRU, and, MRU. Query retrieval times using the Lucene index are an order of magnitude slower than pure memory solutions.

Very Large Graphs - v0.2

In this version of the very large graph implementation, a major update from v0.1 is support for SPARQL queries on the extended model. Furthermore we provide sophisticated algorithms from the field of social networks such as degree centrality and inidividual clustering coefficient that outperform the simple memory management algorithms such as FIFO, LIFO, LRU, and, MRU. We also provide a different way of persisting triples to the Lucene triple store that greatly improves query times.

Very Large Graphs - v0.3

In this version of the very large graph implementation, a major update from v0.2 is the ability to perform reasoning over the extended model. We also provide the ability to use a unified model that transitions from the extended in-memory model to the RDB model when a large number of triples are streamed by an end user application.

Very Large Graphs Javadoc

Professors: Dr. Murat Kantarcioglu and Dr. Bhavani Thuraisingham

Student: Vaibhav Khadilkar

Documents and Publications:

1. Vaibhav Khadilkar, Murat Kantarcioglu, Latifur Khan and Bhavani Thuraisingham, “Efficient processing of large RDF streams using memory management algorithms”, In 9^th International Semantic Web Conference, Shanghai, China, November 7-11, 2010 - Paper, Poster