Electrolysis/Release Criteria/Pageload: Difference between revisions

From MozillaWiki
Jump to navigation Jump to search
(minor formatting changes)
m (reformat page)
 
Line 1: Line 1:
Page load time may regress, up to 5%, if there are other improvements to offset.
Page load time may regress, up to 5%, if there are other improvements to offset.


RASCI:
= RASCI =
 
* Responsible: jimm
* Responsible: jimm
* Accountable: RyanVM
* Accountable: RyanVM
Line 8: Line 9:
* Informed: elan, cpeterson, release management
* Informed: elan, cpeterson, release management


Metrics:
= Metrics =


=== talos tp5/tp5o (page load time) ===  
== talos tp5/tp5o (page load time) ==


* e10s comparison validated: mconley
* e10s comparison validated: mconley
Line 18: Line 19:
** Linux: worse by 3%
** Linux: worse by 3%


=== FX_PAGE_LOAD_MS ===
== FX_PAGE_LOAD_MS ==
 
* Metric owner: rvitillo
* Metric owner: rvitillo
* Latest Spark analysis [https://github.com/vitillo/e10s_analyses/blob/master/beta45-withaddons/e10s_experiment.ipynb e10s_experiment.ipynb] indicates that e10s is better than non-e10s. The median of the e10s population decreased by about 50% wrt to the median of the non-e10s population.
* Latest Spark analysis [https://github.com/vitillo/e10s_analyses/blob/master/beta45-withaddons/e10s_experiment.ipynb e10s_experiment.ipynb] indicates that e10s is better than non-e10s. The median of the e10s population decreased by about 50% wrt to the median of the non-e10s population.
** Need to confirm that data is reliable for comparison.
** Need to confirm that data is reliable for comparison.
** e10s shows some bimodality in the data, not sure how relevant that is.
** e10s shows some bimodality in the data, not sure how relevant that is.
= Bugs =


<bugzilla>
<bugzilla>

Latest revision as of 23:42, 26 February 2016

Page load time may regress, up to 5%, if there are other improvements to offset.

RASCI

  • Responsible: jimm
  • Accountable: RyanVM
  • Supporting: jmaher
  • Consulted: rvitillo
  • Informed: elan, cpeterson, release management

Metrics

talos tp5/tp5o (page load time)

  • e10s comparison validated: mconley
  • Current e10s diff:
    • Windows 7/8: better by 2-5%
    • OS X: worse by 1.5% (but very noisy)
    • Linux: worse by 3%

FX_PAGE_LOAD_MS

  • Metric owner: rvitillo
  • Latest Spark analysis e10s_experiment.ipynb indicates that e10s is better than non-e10s. The median of the e10s population decreased by about 50% wrt to the median of the non-e10s population.
    • Need to confirm that data is reliable for comparison.
    • e10s shows some bimodality in the data, not sure how relevant that is.

Bugs

No results.

0 Total; 0 Open (0%); 0 Resolved (0%); 0 Verified (0%);