My understanding with Lucene.NET has always been that Threading is MY problem, not Lucene's. It's readers are threadsafe (from memory), and the writers are if you are a little careful....
Can you simulate the same thing with a dummy result set - maybe remove the call to lucene (but keep your threading) and return the same result over and over, and see what happens as you ramp it up. On Thu, Jan 20, 2011 at 10:38, Chris Woolum <[email protected]> wrote: > I didn't think Lucene has built in threading using a regular > Indexsearcher or Indexreader. Wouldn't this be a problem with threading > in your Lucene implementation? This is just as far as I knew... Somebody > please correct me if I am wrong. > > Chris > > > > -----Original Message----- > From: Frank Yu [mailto:[email protected]] > Sent: Wednesday, January 19, 2011 1:16 PM > To: [email protected] > Subject: Lucene.Net 2.9 Uses Single Processor under Stress > > Hi All, > > I met a strange situation that the Lucene.Net used only 1 processor on a > quad dual-core processors machine when the search load exceeded a > certain threshold (i.e., 35 searches per second). It took a few minutes > before the load got distributed over all processors. When the load was > below the threshold (i.e., 20 searches per second), all processors got > used evenly. > > The used Lucene.net is v2.9 and the index size is about 4GB with 15 > millions of documents. There was no index update and the load was search > only without any index update. > > Has anyone seen this kind of behavior before? > > Thanks a lot, > > Frank > > > -- Nic Wise t. +64 21 676 418 | @fastchicken | http://www.linkedin.com/in/nicwise b. http://www.fastchicken.co.nz/ | http://www.flickr.com/photos/nicwise mobileAgent (for FreeAgent): get your accounts in your pocket. http://goo.gl/IuBU Note: We are in New Zealand from Dec 2010 to 19th Feb 2011. UK number will work again after that (+44 7788 592 806)
