Skip to content
This repository has been archived by the owner on Feb 20, 2023. It is now read-only.

Fenix startup time increases with the number of open tabs #7035

Closed
hawkinsw opened this issue Dec 7, 2019 · 4 comments
Closed

Fenix startup time increases with the number of open tabs #7035

hawkinsw opened this issue Dec 7, 2019 · 4 comments
Assignees
Labels
P2 Upcoming release performance Possible performance wins

Comments

@hawkinsw
Copy link
Contributor

hawkinsw commented Dec 7, 2019

Why/User Benefit/User Problem

The more tabs that a user carries over between browser sessions, the longer it takes to start Fenix. As evidence of this behavior, consider the following two graphs:

launchtimes2
launchtimes

Along the y axis are start times and along the x axis are the number of open tabs restored from a previous browser session.

Impact

For a user who keeps a large number of open tabs, this means an unacceptable user experience.

Acceptance Criteria (how do I know when I’m done?)

Startup time will remain constant no matter how many open tabs are carried over between sessions.

┆Issue is synchronized with this Jira Task

@mcomella mcomella transferred this issue from mozilla-mobile/perf-frontend-issues Dec 9, 2019
@mcomella mcomella added the performance Possible performance wins label Dec 9, 2019
@boek boek added the P2 Upcoming release label Dec 19, 2019
@mcomella
Copy link
Contributor

Triage: we need to know how many tabs users keep open to prioritize. @ecsmyth is getting numbers for this in #7191 (comment)

@mcomella
Copy link
Contributor

mcomella commented Feb 24, 2020

Action item: determine how bad this is on the G5 (e.g. by taking a run with no tabs, 10 tabs, 20 tabs, 50 tabs, 100 tabs and comparing). If it's bad, consult team (ecsmyth?) and prioritize accordingly.

@mcomella
Copy link
Contributor

Note: tabs are not currently instrumented in fenix startup so, assuming it has no impact on Displayed time, we may not be able to measure until #7781 lands – in our experience, top sites will always load after tabs so by instrumented top sites, we'll be able to observe the difference on tabs from "fully drawn" time

@MarcLeclair MarcLeclair self-assigned this Mar 4, 2020
@MarcLeclair
Copy link
Contributor

As of early March, with all the changes we've had, this is what I've noticed on a G5+ with 10 runs for each # of tabs:
image

As of now, I think we should focus on the weird spike that can be seen when the first tab is opened.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
P2 Upcoming release performance Possible performance wins
Projects
None yet
Development

No branches or pull requests

5 participants