Hi all, As part of an Univ. project, we developed a small architecture: - Clients=N (JMS producers) sends query requests - ActiveMQ Topics L=queue size P=# of queues - Processors=M (JMS clients) execute and sends query results - PostgresQL
were we would vary these variables and observe the throughput. We use the standard benchmark TPCH schema. Although our benchmarks were not comprenhensive we noticed that having fixed relatively large N and M, increasing the P (# queues) the throughput would degrade. We used two different strategies to publish messages: random and round robin. The optimal number of queues P was always found to be one. Any ideas why is that so? Here you can find all the code, slides and (gnu) plots of our experiments: http://dl.getdropbox.com/u/2675033/milestone02_3806.tar.gz Thanks in advance, best regards, Giovanni -- View this message in context: http://www.nabble.com/increasing-number-of-queues-degrades-performance--tp26109931p26109931.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.
