Just out of curiosity, in your Win7 systems, did you change the "MAXCONNECTIONSPERSERVER" and "MAXCONNECTIONSPER1_0SERVER" settings in the registry?
My experience has been that on all Windows systems - this setting throttles the number simultaneous out-bound connections you can make to the same server. I've increased mine to allow 'java.exe' and 'javaw.exe' to go up to 2048 simultaneous connections - and that does seem to make a difference in JMeter behavior. But I can't actually quantify whether it allows all 2048 connections, or if there is some Windows imposed 'hard limit' at 48 or something... -- Robin D. Wilson Sr. Director of Web Development KingsIsle Entertainment, Inc. VOICE: 512-777-1861 www.KingsIsle.com -----Original Message----- From: James Carr [mailto:[email protected]] Sent: Thursday, February 16, 2012 2:02 PM To: JMeter Users List Subject: Complete difference in results when scaling across multiple boxes Hi All, I ran these tests several times, and each time I come back with the similar results. I really don't get it. I am using JDBC sampler with max connections set to 30 for running across 3 machines and 90 when running on one machine. Here's the results: 30 Threads x 3 Machines for 2 minutes Number of Samples: 176,912 Average Response: 62 Std. Deviation: 436.46% 90 Threads x 1 Machine for 2 minutes Number of Samples: 912,954 Average Response: 11 Std. Deviation: 34.14% Machine Specs: RAM: 8GB Processor: Intel Core i7 CPU 870 @ 2.93Ghz System Type: 64-bit Windows 7 Any idea why I see this huge discrepancy? I have double checked on our database on the server side and I verified that we do have the number of connections we expect to see, but I can't understand why the performance looks so lousy with multiple machines. Thanks, James --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
