Parallel session support

[stuartgreen4j]

Currently only a single store session can write to the Neo4j database at any one time. This presents a big challenge for the loading of large triple stores ( >1TB) to Neo4j.

Neo4j is creating a graph and has to perform a MERGE for every triple (lookup the node to check if it exists and attach the property or relationship to it). The inherent lack of order in the RDF file means that the Neo4j transaction must lock the respective node labels & relationships in order to maintain ACID compliance. This prevents parallel import as the process is dependent on the Resource label that is shared by all nodes in the graph.

Investigations ongoing in the use of multiple Resource labels to cover each RDF concept. This may enable the parallelisation of RDF data load to separate subgraphs that can then be knitted together via node merging.