Skip to content

docs(audits): plan io_uring submission-mode benchmark (#1626)#3841

Merged
oferchen merged 1 commit into
masterfrom
docs/iouring-sqpoll-bench-1626
May 7, 2026
Merged

docs(audits): plan io_uring submission-mode benchmark (#1626)#3841
oferchen merged 1 commit into
masterfrom
docs/iouring-sqpoll-bench-1626

Conversation

@oferchen
Copy link
Copy Markdown
Owner

@oferchen oferchen commented May 7, 2026

Summary

  • Adds docs/audits/iouring-sqpoll-bench-plan.md capturing the benchmark plan for comparing regular submit_and_wait(), IORING_SETUP_SQPOLL, and the standard-I/O fallback in crates/fast_io/src/io_uring/.
  • Documents current submission modes, the SQPOLL/CAP_SYS_NICE trade-offs, the proposed workload/queue-depth/privilege matrix, the metrics to collect, and the concrete decision rules for any future default flip.
  • Tracking task Fix platform lookup usage in CLI #1626. No code changes; planning doc only.

Test plan

  • markdownlint docs/audits/iouring-sqpoll-bench-plan.md
  • Verify file is under 350 lines and contains no em-dashes
  • Review for accuracy against crates/fast_io/src/io_uring/config.rs and shared_ring.rs

Locks in the workloads, metrics, and pass/fail criteria for comparing
regular submission, SQPOLL, and standard-I/O fallback in fast_io's
io_uring submission path. Captures current default state, kernel-thread
trade-offs, and the privilege/idle-ms axes the benchmark must vary
before any default flip lands.
@github-actions github-actions Bot added the documentation Improvements or additions to documentation label May 7, 2026
@oferchen oferchen merged commit 40f9049 into master May 7, 2026
8 checks passed
@oferchen oferchen deleted the docs/iouring-sqpoll-bench-1626 branch May 8, 2026 03:50
oferchen added a commit that referenced this pull request May 18, 2026
Locks in the workloads, metrics, and pass/fail criteria for comparing
regular submission, SQPOLL, and standard-I/O fallback in fast_io's
io_uring submission path. Captures current default state, kernel-thread
trade-offs, and the privilege/idle-ms axes the benchmark must vary
before any default flip lands.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant