~/talk

Revisiting Ceph's Performance After 4 Years

Talkat Cephalocon 2024 · 2024-12-04..12-05
links:none on file
slides:none on file

Benchmarks compare Ceph 14.2 → 18.2 on Ubuntu 22.04 using 4K single-thread, QD1 workloads (60s runs) on three 1U servers with 6 OSDs (25 Gb, jumbo frames, BGP L3), AMD EPYC Gen4 + NVMe. Latency drops from ~772 µs (2019) to ~672 µs baseline and up to ~40% lower in tuned runs; disabling OSD logging saves ~15 µs and aggregate IOPS roughly double. Bottlenecks are CPU wake/cstate and Ceph OSD code, not the network; the setup is built and produces repeatable results. 4K single-thread writes AMD EPYC Gen4 + NVMe latency down ~40%

AI-written summary · about

  1. 0:00Introduction
  2. 1:34Speaker background
  3. 3:02Defining performance
  4. 4:11Workload importance
  5. 9:34Benchmarking methodology
  6. 14:08Networking infrastructure
  7. 17:34Comparative results
~/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.