I am trying to write a message transformation application using JRuby. This is a web-app to be deployed in Tomcat. When I benchmarked this new app written in JRuby (with some Java components), with the earlier app written in Java I saw a major performance drop when the load (measured in requests per second) on the application is increased. However, when there is very less load (Tomcat is handling only one request at a time), the performance is almost same as that of Java.
To demonstrate this problem I have written a small standalone app. The application adds up 5000 integers (all less than 1001) Multiple threads are spawned and each thread performs the addition a few number of times As the number of threads are increased, the performance difference between the JRuby and Java version keeps getting larger. The results are same even after using the server mode with sufficient warmup Download the app to run and look at the times. Source is included http://www.nabble.com/file/p21053077/StandAloneXformation.zip StandAloneXformation.zip I want to know whether this kind of performance degradation is known/expected. A few pointers on what can be done to bring the performance close to Java even at higher loads would be a great help. The results on my Win32 machine are follows: Threads JRuby Java ======================= 5 28.18 21.50 10 59.24 22.20 15 125.10 22.92 20 155.54 23.42 25 194.19 28.68 These are average times in milliseconds -- View this message in context: http://www.nabble.com/Major-performance-slowdown-while-multithreading-%28Java-Vs-JRuby%29-tp21053077p21053077.html Sent from the JRuby - Dev mailing list archive at Nabble.com. --------------------------------------------------------------------- To unsubscribe from this list, please visit: http://xircles.codehaus.org/manage_email
