Facebook HTTP/2 and HTTP/3 perf regressed almost 100% on http3.arewefastyet.com tests between October 5-15
Categories
(Core :: Networking: HTTP, defect, P2)
Tracking
()
Tracking | Status | |
---|---|---|
firefox84 | - | --- |
People
(Reporter: cpeterson, Unassigned)
References
(Blocks 1 open bug)
Details
(Keywords: perf, regression, regressionwindow-wanted, Whiteboard: [necko-triaged])
[Tracking Requested - why for this release]:
Jens, looks like Facebook HTTP/2 and HTTP/3 perf regressed almost 100% on http3.arewefastyet.com tests between October 5-15 (Fx83 Nightly), from about 600 ms to 1200 ms:
https://http3.arewefastyet.com/linux64/overview?numDays=60
Neither Cloudflare nor Google regressed during that same period. We should ask Facebook if they rolled out some new server change between October 5-15.
Do we run the http3.arewefastyet.com tests on Firefox Beta and Release builds? Then we would know whether this regression in Fx83 Nightly or a Facebook server change affecting all Firefox versions.
Here are the Networking bugs that were fixed in Fx83 Nightly between October 1-7:
Permalink to Perfherder results for Facebook pause=10s:
https://treeherder.mozilla.org/perf.html#/graphs?series=mozilla-central,2635387,1,13&timerange=5184000&series=mozilla-central,2635350,1,13
Permalink to Perfherder results for Facebook pause=25s:
https://treeherder.mozilla.org/perf.html#/graphs?series=mozilla-central,2635554,1,13&timerange=5184000&series=mozilla-central,2635548,1,13
Permalink to Perfherder results for Facebook pause=35s:
https://treeherder.mozilla.org/perf.html#/graphs?series=mozilla-central,2635766,1,13&timerange=5184000&series=mozilla-central,2635772,1,13
Permalink to Perfherder results for Facebook pause=60s:
https://treeherder.mozilla.org/perf.html#/graphs?series=mozilla-central,2636090,1,13&timerange=5184000&series=mozilla-central,2636127,1,13
Updated•4 years ago
|
Updated•4 years ago
|
Updated•4 years ago
|
Updated•4 years ago
|
Comment 1•4 years ago
|
||
It looks like the website has changed.
This graph shows the number of resources that is loaded:
https://treeherder.mozilla.org/perfherder/graphs?highlightAlerts=1&series=mozilla-central,2694535,1,13&series=mozilla-central,2694500,1,13&timerange=5184000
(the graph says it is ms, but it is not it is gust a count of loaded resources)
The count of resources that is loaded has increased, therefore the increase in load time.
Updated•4 years ago
|
Reporter | ||
Comment 3•4 years ago
|
||
(In reply to Dragana Damjanovic [:dragana] from comment #1)
It looks like the website has changed.
...
The count of resources that is loaded has increased, therefore the increase in load time.
In that case, should we close this bug as invalid?
Comment 4•4 years ago
|
||
I did a double check:
This is a run from September 24th - link:
facebook-scroll domComplete opt 60s: 1097.4
and count is 103
I retriggered the same run on November 3rd - link:
facebook-scroll domComplete opt 60s: 1852.1
and count is 213
This confirms that the web site has changed.
We can close this bug
Updated•4 years ago
|
Description
•