TaskForces/CommunityProjects/LinkingOpenData/OLTPBenchmarks - W3C Wiki (original) (raw)
OLTP Benchmarks for Triplestores
This page is the home of a community effort to define OLTP benchmarks for triplestores. See this message for background.
This effort is part of the Linking Open Data Project
A list of existing RDF store benchmarks if found here.
Overview
The goal of this effort is to define a standard set of OLTP benchmarks that can be run against any triplestore in order to understand its performance in write-heavy environments.
Deliverables
- Document with tests + dataset description
- Reference program for running tests
Roadmap
- Define questions to be addressed by tests
- Use above to define initial set of tests
- Define synthetic, reference dataset
- Draft document
- Publish document (where?)
Questions To Be Addressed
The following questions should inform the development of the tests.
Adds
- How quickly can I load a big batch of triples?
- What size transactions work well for this?
- How is add throughput affected by the size of the triplestore?
- If delayed indexing is possible, is it beneficial to do so for bulk loads?
Updates
- Once populated, how quickly can I write batches of updates?
- Small/medium/big batches?
- How is update throughput affected by the size of the triplestore?
Deletes
- How quickly can I delete a big batch of triples?
- What size transactions work well for this?
- How is delete throughput affected by the size of the triplestore?
Tests
TBD.
Dataset Description
TBD.