~/talk

Scalio: Scaling up DPU-based JBOF Key-value Store with NVMe-oF Target Offload

on YouTube 15m
Talkat OSDI '25 · 2025-07-07..07-09· [networking]

speakers

name:Xun Sun
links:none on file
name:Mingxing Zhang
links:none on file
name:Yingdi Shan
links:none on file
name:Kang Chen
links:none on file
name:Jinlei Jiang
links:none on file
name:Yongwei Wu
links:none on file
name:Quan Cheng Laboratory
links:none on file

resources

video:https://youtu.be/BYzXU2VKbvg
slides:none on file

summary

Scalio is a DPU-based JBOF key-value store that bypasses weak ARM cores by using NVMe-oF Target Offload for SSD I/O and one-sided RDMA for cache access, adds batched writes to reduce issued IO, and enforces linearizability with an RDMA-based cache consistency protocol that marks cache slots with a two-bit occupied/complete state to represent an “incomplete” fill. Implemented on a node with seven Samsung SSDs and 8-core DPUs, Scalio sustains up to 3.3× higher throughput than prior DPU systems across YCSB workloads while fully utilizing SSD bandwidth.

AI-written summary · about

chapters (6)

  1. 0:00Scalio introduction
  2. 2:13Scaling DPU-based JBOF
  3. 4:22NVMe-oF offload
  4. 6:53Linearizability without CPU
  5. 11:10Evaluation setup
  6. 14:45Summary

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.