From: Antoni M. <ant...@gm...> - 2008-10-03 08:41:13
|
Prabhat Singh pisze: > Hi All, > Did we do any kind of performance test on aperture? I tried running > aperture in multi-threaded model on a quadcore processor (3 gz), 16 > GB RAM and tried runing 50 to 200 threads. > > These best performance I saw was 3.5 Megabyes per sec. Did someone else > also tried similar experiment? If yes, can you please share the results > or suggest pointers to increase the performance. > > Thanks, > > -PRabhat > That's very interesting. Could you elaborate a bit more on your test. What crawlers did you run. What was the characteristics of the data sources (few big objects, or thousands small) which data store did you use. Did you have crawlers store data directly in the persistent data store, or did you use temporary in-memory models. How did you measure the throughput, just by looking at the size of the NativeStore after a crawl and dividing it by the time it took, or did you use some more elaborate methods? I assume you're using the alpha version of Aperture you wrote about a while ago. Is it so? There are installations of Aduna Metadata Server that make really heavy use of Aperture crawlers, but I don't know any figures. Antoni Mylka ant...@gm... |