914909 - Browser App First Time Launch Latency

Status: RESOLVED WORKSFORME

(Firefox OS Graveyard :: Gaia::Browser, defect, P1)

Description

[Sandip Kamat]

Based on partner expectations, First time launch latency for Browser app is 1500ms. The bug is to track the user story for 1.2 releases. 1.2 should cause no regression from 1.1.

Comment 1

[Lukas Blakk [:lsblakk] use ?needinfo]

Faramarz - you're engineering mgr for browser so starting this with you to find a proper owner.

Comment 2

[ffos-browser]

A Pivotal Tracker story has been created for this Bug: http://www.pivotaltracker.com/story/show/57244930

Comment 3

[ffos-browser]

Karen Rudnitski deleted the linked story in Pivotal Tracker

Comment 4

[Karen Rudnitski [:kar]]

Is this an actual bug? Has someone actually flagged a regression from 1.1?

Comment 5

[Karen Rudnitski [:kar]]

Sandip - re-reading this original bug makes me feel that this is not complete. The following information is required:
- how are partners coming up with this definition? There is no standard way of measuring 'startup' time. For setting appropriate expectations, can you confirm what your definition of startup time is?
- also note: there is currently no tool that measures our own startup time. There are projects in the works to try and instrument it via our eideticker tool, but as far as I know, this has not been done yet. This would be a requirement and the only way we could measure regressions.
- For our own expectation setting, it would be beneficial for us to measure our startup times based on our definition via our tools and track regressions. But it must be noted that there is no industry-standard way of measuring startups (as far as I'm aware).

Therefore for this bug to be relevant, more info is needed - plus a dependency to create startup measurement tests (which is unlikely to happen for 1.2 :) ).

Comment 6

[Mike Lee [:mlee]]

Qualcomm Profiling results from 2013.09.05 show Browser at

    1452ms for cold launch; up from 1156ms in 1.1
     871ms for warm launch; up from  354ms in 1.1

Warm (subsequent) launch is up ~300ms.

Comment 7

[Mike Lee [:mlee]]

Correction: Warm (subsequent) launch up nearly 500ms.

Comment 8

[Dietrich Ayala (:dietrich)]

Waiting on test results from partner, but assigning to Ben for now to shepherd this.

Comment 9

[Sandip Kamat]

New results published on b2g-release-drivers due to partner confidentiality agreements.

Comment 10

[Ben Francis [:benfrancis]]

According to Datazilla the cold startup time of the browser app is consistently comfortably below 1500ms.

Closing as WORKSFORME. Please re-open if a reproduceable performance problem is identified and provide steps to reproduce.

Thanks

Comment 11

[Ben Kelly [:bkelly, not reviewing]]

(In reply to Ben Francis [:benfrancis] from comment #10)
> According to Datazilla the cold startup time of the browser app is
> consistently comfortably below 1500ms.
> 
> Closing as WORKSFORME. Please re-open if a reproduceable performance problem
> is identified and provide steps to reproduce.
> 
> Thanks

Ben, just FYI, I don't think the datazilla cold_load_time is equivalent with the times Sandip is referencing in comment 0.  The cold_load_time only measures until the mozbrowserloadend event is fired while the partners are looking until the point where first usable screen is painted.

Unfortunately the best way we have to measure this right now is to do a visual examination using a camera.  William Lachance and the a-team have some high speed cameras hooked up to eideticker.  They may be able to run some tests for browser.  I've locally been doing a less precise measurement for contacts app using the 30fps camera in my iphone 4s.

Hope that helps.

Comment 12

[Ben Francis [:benfrancis]]

Thanks for the information bkelly,

We can't block the release on a koi+ bug that may or may not actually be a problem and has no steps to reproduce that engineering can follow. So please do re-open the bug if we identify that there actually is an issue with startup time and that some engineer somewhere has a way to reproduce it.

Trying to drive down those blockers!