~/talk

Global Capacity Management With Flux

Talkat OSDI 2023 · 2023-07-10..07-12· [distributed-systems, networking]
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
links:none on file
name:Jeff Qin
links:none on file
links:none on file
slides:none on file

Flux automates regional capacity allocation using RPC tracing to build per-service capacity models, then runs a joint capacity and traffic solver that optimizes placement across tens of regions with heterogeneous hardware. Plans span thousands of services and tens of products, sometimes moving up to 300,000 servers per plan and operating on millions of servers overall. A sequenced capacity orchestrator safely enacts plans, reduces DR buffer needs and capex, and has run nearly autonomously in production for three years.

AI-written summary · about

  1. 0:00Global capacity management intro
  2. 1:30Regional capacity challenges
  3. 3:16Service dependency graphs
  4. 5:21Heterogeneous hardware issues
  5. 7:27Capacity management Gordian knot
  6. 8:30Manual processes failed at Meta
  7. 9:53Introducing Flux and pillars
  8. 11:17Service modeling with RPC tracing
  9. 12:21Traffic and capacity optimization
  10. 13:24Safe capacity orchestration
  11. 14:27Flux's 3-year impact
  12. 15:31Flux capabilities summary
~/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.