Skip to content
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
This repository was archived by the owner on Nov 21, 2026. It is now read-only.

Benchmarks: 2026-11 is the only instrument on the measure side #428

Closed

Description

@tae898

DECISIONS #196 (2026-10-21): "just remove the old method from dbbench. also remove the old data. let's go for November directly without October" and "october was dev". The harness retires the 2026-10 and September instruments, so 2026-11 is the only instrument on the measure side. This is the measure-side part of the umbrella #410 (one stage generator and the publish side are separate PRs).

What changes: BENCH_INSTRUMENT defaults to 2026-11 and any other value is refused with a message (the stamp on every row and the selector shape stay, so a later instrument can be added). The OFF code path of every switch that #413 made "on under 2026-11" is deleted: sparse warm-up, CPU idle hold, dense id index, embedded -Xms, large-result Arrow path, official driver as the served headline with the lean client as the sensitivity twin. The lean2 client, the requests client path, test_lean_http2.py and the _under_2026_10 tests go. The runner forwards BENCH_INSTRUMENT to every cell from its own value. CAMPAIGN, PROTOCOL, FAIRNESS, COMPARATORS, PAGE-SPEC and the bindings benchmark pages describe 2026-11 as the method.

Env overrides kept only where PROTOCOL.md section 7 documents them as numeric overrides: BENCH_SPARSE_WARMUP=N and BENCH_CPU_DMA_LATENCY_US=N. Nothing can turn a method choice off.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions