Moving RDF Stream Processing to the Client
0 likes352 views
Stream-processing SPARQL endpoints hosted on web servers are expensive due to an unknown number of clients, unbounded query complexity, and the server doing all the work while clients wait for results. Publishing dynamic data with Triple Pattern Fragments and making clients contribute more to the processing addresses this by annotating triples with timestamps, having clients re-evaluate queries as needed based on the timestamps, and designing the server interface to handle simple requests while putting most of the work on clients.
Moving RDF Stream Processing to the Client
- 1. Moving RDF Stream Processing to the Client Stream-processing SPARQL endpoints are expensive to host on the Web Server does all the work, clients wait for results Unknown number of clients Unbounded query complexity Publish dynamic data with Triple Pattern Fragments and make clients work for their results [email protected] @rubensworks ... ... Clients want data, so they should contribute to the cost All requests are cacheable <s> <p> <o> 15:16 GMT+2 23:42 GMT+2 Triples are annotated with time Clients re-evaluate queries at the time they are needed Data-driven update frequency Server exposes interface handling simple requests, clients do most of the work