An Evaluation of Triple-Store Technologies for Large Data Stores

Rohloff, Dean, Emmons, Ryder, Sumner

rdf semantic web triple store evaluation query

@inproceedings{rohloff:otm-2007,
  title={An Evaluation of Triple-Store Technologies for Large Data Stores},
  author={Rohloff, K. and Dean, M. and Emmons, I. and
           Ryder, D. and Sumner, J.},
  booktitle={On the Move to Meaningful Internet Systems 2007: OTM 2007 Workshops},
  pages={1105--1114},
  year={2007},
  organization={Springer}
}

Looks at Sesame, Jena, and Allegro in working with large datasets

Primarily LUBM datasets, mentions University Ontology Benchmark

Metrics: Load time, repository size, response time

• Cumulative load time: How long to long data and ontologies (hours!)
• Query response time: Mean of execution time over four identical queries
• Completeness: Complete iff returns all correct responses
• Soundness: Sound iff only returns correct responses
• Disk-Space usage: Amount of disk space used to load data and ontologies

Multiple styles of query used

• Low volume, low complexity (Lehigh query 1)
• High volume, low complexity (Lehigh query 2)
• High complexity (Lehigh query 9)
• High volume: Large portion of data returned in result set
• Complexity: Substantial processing time required

Queries manually ported for different input languages