Last modified 4 years ago Last modified on 26/02/10 13:59:26

5store

5store is the name for the latest clustered RDF store developed by Garlik.

It's external behavior is similar to 4store's (similarly named tools, HTTP server etc.) but it removes many of the restrictions of 4store, is more efficient, and much more scalable.

Key differences are:

  • No effective limit on the number of triples stored. [1]
  • Import performance scales with cluster size, we've seen average import speeds over 700kT/s on a 10 node cluster. [2]
  • Designed for clusters of up to 1000 machines.
  • Very efficient storage, approaching twice the triples/machine density of 4store.
  • Command line admin tools and HTTP server can work at the same time.
  • Very efficient streaming of query results, better handling of large result sets.
  • Can delete arbitrary triples efficiently.

[1] 1TT should be easily achievable, given a large enough cluster.
[2] 1MT/s is about the practical limit on gigabit Ethernet, to go beyond that you'd need 10gigE or similar.

Licensing

5store is now backing Garlik's production systems. If you have need of an RDF store of this scale then it may be possible to license a copy from Garlik, please contact us for more information.

TSO have licensed 5store to host government datasets. There's more information in their press release.