~/talk

Speculative Recovery: Cheap, Highly Available Fault Tolerance with Disaggregated...

on YouTube 19m
Talkat USENIX ATC '22 · 2022-07-11..07-13· [distributed-systems]

speakers

name:Nanqinqin Li
links:none on file
name:Nanqinqin Li
links:none on file
name:Anja Kalaba
links:none on file
name:Michael J. Freedman
links:none on file
name:Wyatt Lloyd
links:none on file
name:Amit Levy
links:none on file

resources

video:https://youtu.be/7BBEfvNzDF0
slides:none on file

summary

Speculative recovery claims to cut REDS failover time by running failure detection and recovery in parallel using two new primitives super and collapse, implemented as a fast co‑located copy‑on‑write clone in a prototype called specReds. Evaluation on Docker+NVMe with MySQL/Postgres/MariaDB shows existing clone implementations raise write latency ~30×, while the co‑located clone matches regular disk performance and lets recovery race the primary. In a false‑positive test (primary returns after 10s, typical timeout 5s causing a 1‑minute recovery), speculative recovery aborts unnecessary failovers and significantly reduces outage time.

AI-written summary · about

chapters (7)

  1. 0:00Start
  2. 1:26A nascent fault-tolerance technique
  3. 11:33super uses copy-on-write
  4. 12:32collapse uses a dirty bit to monitor writes
  5. 13:17Three components of a speculative recovery system
  6. 14:15The experiment setup
  7. 15:11Application recovery latency from various disk states

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.