1271321 - Measure the max number of windows open concurrently per "session fragment"

Status: RESOLVED DUPLICATE of bug 1271304

(Toolkit :: Telemetry, defect, P3)

Description

[Alessio Placitelli [:Dexter]]

Count the max number of windows opened in a session. Ideally, this should come after bug 1271319 (which has some more context).

Context: https://docs.google.com/spreadsheets/d/1G33YEBL2-hSaF0-HrHPi0HVwANoTpYHsnevmMd1eZ1U/edit#gid=0

Comment 1

[Benjamin Smedberg]

Did you see my recommendation in the doc? I think we should record, at the time each window is opened, the number of windows currently open. That will solve both this and 1271319 together without having to worry about state. You can see the total number of windows opened by counting the measurements, the largest number of windows open by finding the highest recorded value, and you also get some interesting distribution metrics.

Comment 2

[Georg Fritzsche [:gfritzsche]]

Sadly the conversation is a bit distributed here between these two documents:
* wishlist: https://docs.google.com/spreadsheets/d/1G33YEBL2-hSaF0-HrHPi0HVwANoTpYHsnevmMd1eZ1U/edit#gid=0
* PRD: https://docs.google.com/document/d/1UuztyM-fdQBR88QjOiMIbqlHhLGMVfvPF_xbyhKWmOQ/edit?ts=56c378e3#heading=h.ygjz8s3u0xtp

I've commented on the PRD conversation & i wonder if we really need those detailed metrics here.
In summary:
* this would require further work to make the max and open counts available (for analysis or dashboards like t.m.o)
* this means that we have to look at assembled client sessions, as not all individual subsessions would have the data

Comment 3

[Benjamin Smedberg]

What "further work" are you worried about? Base t.m.o already won't solve this problem because it's not using session fragments, but it would be trivial to aggregate max(histogram).

In terms of the ping-assembly, though, are you planning on recording this in every subsession whether or not that subsession opens any windows? That surprises me a bit because it would involve a fair bit of extra state tracking. I figured that anyone looking at this would be primarily looking at it by user anyway.

Comment 4

[Georg Fritzsche [:gfritzsche]]

So, with the proposed approach, for all the measurements of window & tab open events and window & tab max open counts we'd need to assemble the subsessions per-client in order.
Looking at individual subsessions is not possible and there seem to be lots of ways to look at this data the wrong way.

I currently prefer to record the state in each subsession.
Having measures in each subsession seems a lot simpler to look at and keeping track of the states for the limited measures here is relatively easy.

Comment 5

[Georg Fritzsche [:gfritzsche]]

(In reply to Georg Fritzsche [:gfritzsche] from comment #4)
> I currently prefer to record the state in each subsession.
> Having measures in each subsession seems a lot simpler to look at and
> keeping track of the states for the limited measures here is relatively easy.

We'll move forward with this at it would most directly cover the questions being asked (correlating other data in a subsession with this).
The state tracking client-side is not complex in this case.