1328471 - Expand TOTAL_SCROLL_Y histogram and make it exponential

Status: RESOLVED WONTFIX

(Toolkit :: Performance Monitoring, defect)

Description

[:Harald Kirschner :digitarald]

Right now TOTAL_SCROLL_Y is set up with "kind": "count" and default settings for "buckets" and "high".

I would try settings similar to PAGE_MAX_SCROLL_Y but with a higher max as the scroll amount is accumulated throughout a whole session.

"kind": "exponential",
"high": 1000000,
"n_buckets": 50,

This is a best guess as we can not tell the better "high" setting from the Nightly probe data.

Comment 1

[:Harald Kirschner :digitarald]

mreid recommended to land this fix with a renamed probe to avoid side effects in the telemetry aggregator and overlapping old/new data.

To avoid appending versions a possible new name could be TOTAL_SCROLL_Y_EXP.

Comment 2

[Ryan Hunt [:rhunt] (on leave until early May)]

For TOTAL_SCROLL_Y, we use the "count" histogram because it allows us to call Accumulate once each scroll event, and it will handle totaling them up and giving one value for each session fragment.

From an irc discussion, it didn't sound like exponential would give us that behavior? It can be worked in if it doesn't though.

Comment 3

[Georg Fritzsche [:gfritzsche]]

(In reply to Ryan Hunt [:rhunt] from comment #2)
> From an irc discussion, it didn't sound like exponential would give us that
> behavior? It can be worked in if it doesn't though.

Right, with exponential you would have to accumulate locally and record it when the life-time of the frame/document/... ends.

Comment 4

[:Harald Kirschner :digitarald]

Putting this bug on hold for now. The limitation of buckets in count histograms is in the aggregator, so the raw data contains the right counts.

The fix and backfilling is tracked in bug 1328678.