Performance metrics Round 2 for GeckoView streaming
Categories
(Core :: Performance: General, enhancement, P1)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox75 | --- | affected |
People
(Reporter: bdekoz, Assigned: bdekoz)
References
(Depends on 2 open bugs)
Details
Attachments
(1 file, 1 obsolete file)
|
4.25 KB,
text/plain
|
Details |
Exfiltrate an additional subset subset of existing telemetry probes for adequately defining page load from gecko telemetry to fenix/android using the new glean/geckoview_streaming approach. This fills in holes and builds on first set of metrics that landed in BZ 1580077.
The known holes in existing datasets are:
Network
network page weight/total bytes downloaded
time to first network idle
redirects
reloads/redirect
Startup/Navigation
cold warm startup
F/B cache
Memory
system memory total
current resident set size for Mozilla apps
Environment
system cpu count
system cpu speed
system memory size
system disk cache size
Potentially useful metrics, but not necessarily a requirement for this round of work:
Resources (layout and necko)
how many images
how many flex boxes
how many reflows
if/how was stuff pushed out of cache (css forced out by video)
FXA sync
time to migration for
passwords
bookmarks
history
how many tabs are sent from desktop for mobile
Media playback
resolution of video file (not a fixed number, can change)
has resolution changed within playback
average frame rage
size
playback minutes
if CDN?
“time to transfer from CDN”
stalls, want to decode video and audio but don’t have data yet
if user initiated or autoplay
if subtitiles
New metrics relative to responseStart
relPageLoadTime
relDomInteractiveTime
relDomContentLoadedTime
| Assignee | ||
Updated•6 years ago
|
Updated•6 years ago
|
Comment 1•6 years ago
|
||
For pageload, I think it would be valuable to know where the navigation originated from.
e.g.
-cold process applink (i.e. the gecko process had been killed by the OS or user)
-warm process applink (gecko process was already running)
-url bar entry
-link selection from document
-history or similar selection
-refresh
-custom tabs
-?
| Assignee | ||
Comment 2•6 years ago
|
||
| Assignee | ||
Comment 3•6 years ago
|
||
Feedback from round one is that network was mostly right, Dragana suggested these two addtions. The real issue is that there is no engagment info, so fixing this is the focus of this round.
| Assignee | ||
Comment 4•6 years ago
|
||
Comment 5•6 years ago
|
||
| Assignee | ||
Comment 6•6 years ago
|
||
Agh, hang on with this patch, sorry double checking with data sci but delayed.
Comment 7•6 years ago
|
||
Triage: Benjamin, is this still needed? If not, we were wondering if this should be closed.
Updated•4 years ago
|
| Assignee | ||
Comment 9•3 years ago
|
||
Closing, metric shaping went other ways.
Updated•1 year ago
|
Description
•