higher drain rate with larger ep queue size in performance graphs.

From Sharon:

1. The ep queue size graph does not make sense. it shows much higher queue size for 2.0, while the drain rate is also much higher.
Can you verify that you are not mixing the results from the builds?

From Ronnie:

#1> The results are mixed. Plz see the following graph from vms, which shows similar behavior.

Here is my guess:

From the graph, 2.0 has more divergences in the draining rate (even the avg in 2.0 is better) , which build up the queue. In other words, from a blackbox perspective, 1.8 seems to flush more often.

From Steve:

@Ronnie Sun i think #1 (higher queue sizes in 2.0 compared to 1.8) deserves a jira bug – work with ep-engine guys towards at least an explanation?