HttpClient - use ConcurrentHashMap instead of HashMap?
AbstractCookieSpec uses an unsynchronised HashMap, so is not thread-safe. AIUI, ConcurrentHashMap has much better performance than a synchronised HashMap - but of course we are not using synch here. A quick test using the code at [1] (1 thread, 1 VM) appears to show that the performance of CHM and HM is fairly similar: Map type: HashMap Runtime:3000 Number of threads: 1 Write probability: 0.6 Remove probability: 0.02 Ops per second: 1260837.0 Map type: ConcurrentHashMap Runtime:3000 Number of threads: 1 Write probability: 0.6 Remove probability: 0.02 Ops per second: 1313171.0 So I wonder if we should consider replacing HashMap with ConcurrentHashMap in Cookies and elsewhere in HttpClient? [1] http://unserializableone.blogspot.com/2007/04/performance-comparision-between.html - To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
Re: HttpClient - use ConcurrentHashMap instead of HashMap?
Why did you post this to the commons dev list instead of http-components dev? Ralph On Jan 9, 2011, at 7:32 AM, sebb wrote: AbstractCookieSpec uses an unsynchronised HashMap, so is not thread-safe. AIUI, ConcurrentHashMap has much better performance than a synchronised HashMap - but of course we are not using synch here. A quick test using the code at [1] (1 thread, 1 VM) appears to show that the performance of CHM and HM is fairly similar: Map type: HashMap Runtime:3000 Number of threads: 1 Write probability: 0.6 Remove probability: 0.02 Ops per second: 1260837.0 Map type: ConcurrentHashMap Runtime:3000 Number of threads: 1 Write probability: 0.6 Remove probability: 0.02 Ops per second: 1313171.0 So I wonder if we should consider replacing HashMap with ConcurrentHashMap in Cookies and elsewhere in HttpClient? [1] http://unserializableone.blogspot.com/2007/04/performance-comparision-between.html - To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org - To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
Re: HttpClient - use ConcurrentHashMap instead of HashMap?
On 9 January 2011 16:20, Ralph Goers ralph.go...@dslextreme.com wrote: Why did you post this to the commons dev list instead of http-components dev? Oops, sorry! However, the timings might still be interesting for Commons components... Ralph On Jan 9, 2011, at 7:32 AM, sebb wrote: AbstractCookieSpec uses an unsynchronised HashMap, so is not thread-safe. AIUI, ConcurrentHashMap has much better performance than a synchronised HashMap - but of course we are not using synch here. A quick test using the code at [1] (1 thread, 1 VM) appears to show that the performance of CHM and HM is fairly similar: Map type: HashMap Runtime: 3000 Number of threads: 1 Write probability: 0.6 Remove probability: 0.02 Ops per second: 1260837.0 Map type: ConcurrentHashMap Runtime: 3000 Number of threads: 1 Write probability: 0.6 Remove probability: 0.02 Ops per second: 1313171.0 So I wonder if we should consider replacing HashMap with ConcurrentHashMap in Cookies and elsewhere in HttpClient? [1] http://unserializableone.blogspot.com/2007/04/performance-comparision-between.html - To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org - To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org - To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
