Skip to content

docs(rqe-optimizer): port sketch-bench doc edits; fold case-studies-6.3.1 into case-studies - #920

Open
milindsrivastava1997 wants to merge 1 commit into
mainfrom
docs/rqe-optimizer-leftovers
Open

milindsrivastava1997 wants to merge 1 commit into
mainfrom
docs/rqe-optimizer-leftovers

Conversation

@milindsrivastava1997

Copy link
Copy Markdown
Contributor

Leftovers from the rqe-optimizer move (#915).

Docs. sketch-bench#234 edited e2e_planning_flow.md and rqe_optimizer_hydra.md after they were copied here. This ports its edits: the asap_sketchlib 0.3.0 mentions, Hydra's reported footprint (Hydra::memory_bytes), and the rows path (sketch-bench/sketch-bench/src/rows/mod.rs).

results/case-studies-6.3.1/ removed. Its two workloads are case-studies/'s base-0 tables. The results match for every method (asap, asap-norollup, perquery, autosketch) under both CPU and Fargate weights. Its per-method tables and query listings now sit under "Every method on the base workloads" in case-studies/README.md. rqe-optimizer/README.md now points at case-studies/.

case-studies/README.md corrections (checked against the result files):

  • The anomaly batch SLA was listed under "Knobs that do not change the plan", but it raises the cost from 0.114 (3 MVs) to 0.377 (6 MVs) at 0.3 ms. It is now finding 5.
  • Costs scale linearly with series count and ingest rate only when there is no SLA.
  • Under Fargate prices with the drill-down at 50 endpoints, the planner picks 2 MVs, not 1.
  • Restored the missing "Caveats" heading, so the CPU-only notes no longer sit under "Fargate prices".

plot_sec63_paper.py. cost_bars was only ever called with "column". Its wide branch would have saved over sec63_cost_bars, which is cost_bars_single's output. The PR removes that branch, renames the function to cost_bars_column and lists every figure in the docstring. I regenerated all six figures from main and from this branch, and the PNGs are byte-identical.

Not in this PR: the sketch-bench#159/#172/#189 references in rqe-optimizer/src stay as they are until those issues are moved here.

🤖 Generated with Claude Code

https://claude.ai/code/session_014qhwq8JRnKRVPR15fxLXyq

….3.1 into case-studies

- docs: the asap_sketchlib 0.3.0 mentions, Hydra's footprint and the rows
  path, as sketch-bench#234 left them after the move.
- results: case-studies-6.3.1 duplicated case-studies' base-0 (identical
  cost for every method and weight setting); its per-method tables now live
  in case-studies/README.md.
- case-studies/README.md: the anomaly batch SLA does change the plan
  (0.114 -> 0.377); linear scaling holds only with no SLA; Fargate at 50
  endpoints picks 2 MVs, not 1; restore the Caveats heading.
- plot_sec63_paper.py: drop cost_bars' never-called wide layout (it would
  have overwritten sec63_cost_bars); the docstring lists every figure. All
  figures are byte-identical.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014qhwq8JRnKRVPR15fxLXyq
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant