We read in large chunks of messages at a time and start processing them. Some messages wait longer in internal queues for CPU time to become available to proces

We read in large chunks of messages at a time and start processing them. Some messages wait longer in internal queues for CPU time to become available to process them. The benchmark we use is intentionally very light computationally. We did that because, if we had a heavy computation then the overhead from Wallaroo would be hidden.