From http://blogs.illation.com.au/2009/08/drools-performance-waltzdb/
Post-Run Memory Used
Drools/4 consistently allocates a great deal more memory than drools/5 at the start, and then slowly releases it as the server optimizations settle down. In all these graphs it’s interesting to note that there’s very little or no difference between the memory pre and post run figures. This has been the case in all the tests run to date, so from this point forward we’re just going to use post-run memory.
As we’ve seen in the past, drools/5 allocates less memory up front, and then gyrates for a while before settling.
An overlay of drools/4 and drools/5 memory shows the same pattern as previous tests. This could be related to the elimination of shadow facts in drools/5. In drools/4 a shadow object must be created for each of the facts asserted into working memory, however in all benchmarks run so far drools/4 ultimately uses less overall memory.
Post-Run Memory Used
Drools/4 consistently allocates a great deal more memory than drools/5 at the start, and then slowly releases it as the server optimizations settle down. In all these graphs it’s interesting to note that there’s very little or no difference between the memory pre and post run figures. This has been the case in all the tests run to date, so from this point forward we’re just going to use post-run memory.
As we’ve seen in the past, drools/5 allocates less memory up front, and then gyrates for a while before settling.
An overlay of drools/4 and drools/5 memory shows the same pattern as previous tests. This could be related to the elimination of shadow facts in drools/5. In drools/4 a shadow object must be created for each of the facts asserted into working memory, however in all benchmarks run so far drools/4 ultimately uses less overall memory.
扫一扫
4655
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
