~/talk

Kamino: Efficient VM Allocation at Scale with Latency-Driven Cache-Aware Scheduling

on YouTube 17m
Talkat OSDI '25 · 2025-07-07..07-09· [cloud-infrastructure]

speakers

name:David Domingo
links:none on file
name:Hugo Barbalho
links:none on file
name:Marco Molinaro
links:none on file
name:Kuan Liu
links:none on file
name:Abhisek Pan
links:none on file
name:David Dion
links:none on file
name:Thomas Moscibroda
linkedin:thomas-moscibroda-26a431a4
name:Sudarsun Kannan
links:none on file
name:Ishai Menache
links:none on file

resources

video:https://youtu.be/KoN8KZPIelA
slides:none on file

summary

Kamino implements a latency-driven cache-aware scheduler on top of Protein, adding a multiQ design, request classifiers, a latency estimator, and the LAC cache algorithm that estimates hit/miss times per agent to pick the lowest expected end-to-end latency. Simulation on six high-traffic traces shows 42% reduction in average latencies and up to 55% tail improvement; a production rollout across five Azure zones yields 33% decrease in cache miss rates and 17% lower memory use.

AI-written summary · about

chapters (9)

  1. 0:00VM allocation challenges
  2. 3:07Protein allocator overview
  3. 4:54Memory bottlenecks
  4. 7:44Cache reuse vs load balancing
  5. 9:07Cache-aware latency scheduling
  6. 12:57Hierarchical caches and Lac Cache Rule
  7. 14:22Performance improvements
  8. 15:26Azure deployment results
  9. 16:50Conclusion

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.