~/talk

ATP: In-network Aggregation for Multi-tenant Learning

Talkat NSDI ’21 · 2021-04-12..04-14· [machine-learning, networking]
links:none on file
links:none on file
links:none on file
name:Yixi Chen
links:none on file
name:Wenfei Wu
links:none on file
links:none on file
slides:none on file

Claims up to 38–66% training throughput improvement for multi-tenant distributed DNN jobs by moving gradient aggregation into programmable switches. Implements in-network aggregation in P4 on Barefoot Tofino-style hardware, using dynamic per-packet aggregators hashed by gradient index, two-level TOR+aggregation-switch reduction, and a best-effort fallback that forwards missed aggregations to parameter servers. Ensures exactness via sequence-detect retransmit, converts floats to 32-bit scaled integers to avoid switch FP, and evaluates on a cluster against static partitioning and prior systems.

AI-written summary · about

  1. 0:00Introduction
  2. 0:28Parameter Server Architecture
  3. 3:06Agenda
  4. 3:31Aggregation
  5. 4:41Best Effort
  6. 7:45Additional Challenges
  7. 9:46Evaluation
  8. 10:19Dynamic Allocation
  9. 10:46Results
  10. 11:39Summary
~/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.