Still investigating on the situation where Clerezza doesn'tswllo stress
test. Today it passed the following 10 millions triples challenge:
Setting:
- 100 concurrent threads
- 1000 * 10000 requests
- requested media-type: application/rdf+xml
generating resource method:
@GET
public GraphNode list() {
TripleCollection resultGraph = new SimpleMGraph();
GraphNode result = new GraphNode(new BNode(), resultGraph);
RdfList list = new RdfList(result);
LockableMGraph contentGraph = cgProvider.getContentGraph();
Lock l = contentGraph.getLock().readLock();
l.lock();
try {
Iterator<Triple> greetings = contentGraph.filter(null, RDF.type,
GREETINGS.Greeting);
while (greetings.hasNext()) {
list.add(greetings.next().getSubject());
}
} finally {
l.unlock();
}
return result;
}
Data:
- 2 resources of type GREETINGS.Greeting
Result:
For every block of 10000 requests the following is true:
- 90% of the requests where answered in less than 161 ms
- 50% of all requests where answered in less than 58 ms
- the longest request took between 15687 and 3105 ms
- mean time between 109 and 221 ms
- No requests failed
No rocket science statistics, but it gives an idea.
A next step is to add an ssp.
Reto
