Increased latency with longer run

[uttampawar]

We have noticed an issue of increased latency (and lower throughput) when the workload runs for extended period. This happens only in specific scenario such as when the client issue 'POST' and 'DELETE' requests. While with 'GET' type requests we get consistent throughput (and constant latency)

One theory is that mongodb executes 'indexing' action after every POST and/or DELETE action.
I'll update more data soon. Any ideas?