The "CPU usage across our Elasticsearch cluster" chart shows a very large spread of utilization at almost all times. This is read-only since you are re-writing offline, right? What do you think is driving this variation?

Disclosure: I am VP, Product Strategy at MarkLogic...