~/talk

Alohamora: Reviving HTTP/2 Push and Preload by Adapting Policies On the Fly

on YouTube 12m
Talkat NSDI ’21 · 2021-04-12..04-14· [web-platform, machine-learning]

speakers

name:Nikhil Kansal
links:none on file
name:Murali Ramanujam
links:none on file
name:Ravi Netravali
linkedin:ravi-netravali

resources

video:https://youtu.be/9LGXX_iXuwo
slides:none on file

summary

Alohomora uses reinforcement learning to pick per-load HTTP/2 push and preload actions, trains one model per cluster of structurally similar pages, and evaluates policies with a fast page flow simulator that runs in milliseconds with ~0.4% median error. Training that would take 10 days for a large site drops to ~20 minutes, learned policies speed page loads by ~19–61% and give ~3.6–4× more benefits than prior systems while never degrading performance in evaluated mobile scenarios.

AI-written summary · about

chapters (4)

  1. 0:00Introduction
  2. 0:30Overview
  3. 5:14Challenges
  4. 9:32Implementation

related talks

~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.