Hello, >From http://stackoverflow.com/questions/24413088/storm-max-spout-pending looks like max spout pending is the maximum number messages that are within the topology at any point in time.
From http://www.michael-noll.com/blog/2013/06/21/understanding-storm-internal-message-buffers/ the various topology buffer sizes define the number of messages in the internal buffers. should the max spout pending and internal buffer sizes be modified together to get low latency ? for ex i have a topology with kafka spout and two bolts. the latency in each bolt is 0.048 and 1.8 ms though the overall latency is about 45 ms. Also i reduced max spout pending to 192 to bring down the latency as i think if more messages are waiting in queue then over all latency will be higher, but still the time difference is around 30 times more than the sum of individual bolt latency. regards, anishek