Split random access summaries into hot/cold - #26
Conversation
Signed-off-by: Will Manning <will@willmanning.io>
|
Vercel preview for |
Signed-off-by: Will Manning <will@willmanning.io>
Signed-off-by: Will Manning <will@willmanning.io>
Signed-off-by: Will Manning <will@willmanning.io>
Signed-off-by: Will Manning <will@willmanning.io>
Signed-off-by: Will Manning <will@willmanning.io>
Signed-off-by: Will Manning <will@willmanning.io>
Signed-off-by: Will Manning <will@willmanning.io>
lwwmanning
changed the title
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 615b19c3d4
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| * It is independent of the producer-facing benchmark schema version. | ||
| */ | ||
| const GROUP_PAYLOAD_CACHE_VERSION = 'v7'; | ||
| const GROUP_PAYLOAD_CACHE_VERSION = 'v8'; |
There was a problem hiding this comment.
The random-access chart wire shape now changes from format-only series to format:hot/format:cold, but this bump versions only the group and group-chart caches; chartPayloadCached still uses the unchanged data-cache:chart-payload:n100 key. Because the inspected .github/workflows/web-deploy.yml does not invalidate the Data Cache, default requests to /chart/:slug and /api/chart/:slug can keep returning the pre-deploy format-only payload for the 24-hour backstop, while the newly versioned group bundle already shows split modes. Add a version to the standalone chart cache key as well.
Useful? React with 👍 / 👎.
This branch was successfully deployed
2 participants
Split the Random Access summary into Hot and Cold panels. Each panel ranks formats independently from dataset-grouped results and reports the arithmetic mean across datasets. Hot reuses an accessor after a one-second untimed warm-up. Cold includes file open and metadata parsing in each timed take. It does not clear the OS page cache. Charts label both modes and use dashed lines for Cold results. Write Throughput now reports aggregate values in MB/s, while Scan Throughput remains in GB/s.
This depends on Vortex PR #9631 for the mode definitions and PR #9649 for the
open_modeproducer and ingest path. Migration 010 marks historical rows ascached, and schema version 3 coordinates the deployment.