Hi Everyone, I try to capture the relationship between throughput and latency across production cluster of two worker machines. So the I submitted the topology multiple times with different spouts and bolts. So I noticed the following not clear results to me: 1- For the same number of spouts and bolts (even 1 spout and bolt), the latency is very different from one run to another. For example, in the first run it about 120 ms, in the second run about 3090 ms !!!2- I think that when the parallelism of spouts and bolts increased, the average throughput and latency should be decreased, but it fluctuates and does not show apparent results !!3- I was wondering if there is any work shows the performance of storm with different spouts and bolts.. Hope anyone can help to give us some explanations or more information about above points. Thank you--- Best Regards Walid