ops drop for mixed workload

Please refer to the attached graph.

Ops dropped at 10000 seconds. Cache miss ratio bumped of course. While disk read dropped (iostats). And bg_wait and bg_load suddenly bumped.

From steve:

• was looking through your recent K-V mixed test (Oct 16, 2.0.0-1847)...

http://bit.ly/R4TULi

And I was seeing the ops/second drop. Can you create a jira bug, at least so @Chiyoung Seo and team can take a look and theorize an explanation?

Some relevant pages from the *.pdf – one is ops/sec; one is disk-reads/sec; one is active resident-item-ratio, one is docs actual disk size