We should study the distribution of the ping sizes for the release channel, since most of the users on that channel are "opt-out". We should check Kibana for the basic numbers, compare them to what we except, and move on from there to investigate possible problems.

Mark, can you remind us what average sizes for the opt-out pings your cost calculations were based on?

I assumed an average "base set" ping was 20KB in size.

I was able to setup a Kibana query to compare base set pings vs extended set pings [0]. It looks like the average size of the base data ping is about 10kb (yay!). Base set - min 3,745 B - max 545,368 B - mean 10,477 B - std_deviation 2,890 B Extended set - min 4,703 B - max 1,031,365 B - mean 95,944 B - std_deviation 19,040 B I'll follow up with a distribution of the ping sizes. [0] - https://kibana.shared.us-west-2.prod.mozaws.net/#dashboard/temp/AVC391n833_SLpV1IoRF

An updated dashboard (Kibana 4) with the size distributions: https://goo.gl/1HGmlC

The discussion on FHR-Dev: https://mail.mozilla.org/pipermail/fhr-dev/2015-November/000673.html