Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 2 additions & 3 deletions rqe-optimizer/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -14,8 +14,7 @@ The crate moved here from ProjectASAP/sketch-bench, which still measures the
sketches and exports the tables the runner reads: the saturation study
(`sketch-bench/scripts/study_saturation.py`), the cost table, and the
eval-table exporters (`sketch-bench/scripts/export_autosketch_eval_table.py`,
`sketch-bench/scripts/case_study_tables.py`,
`sketch-bench/scripts/export_case_studies.py`). Paths written
`sketch-bench/scripts/case_study_tables.py`). Paths written
`sketch-bench/...` are in that repo, and commands using `$SKETCH_BENCH` run
from a checkout of it.
`sketch-bench#N` is an issue or PR there.
Expand All @@ -36,7 +35,7 @@ eval on a node, copy the inputs it reads there yourself, e.g.
| `docs/` | the problem statement, candidate and cost-model design, Hydra, and the end-to-end planning flow | |
| `results/autosketch-vs-asap-hydra/` | **the current §6.3 results** (2026-10-11; classic and all rerun 2026-10-13 without templates 7 and 9; all rerun 2026-10-14 on the denser grid, sketch-bench#213): roll-ups, the roll-up ablation, Hydra candidates, the classic, multi-grouping and mixed + multi-grouping workloads, and each plan | current |
| `results/e2e-rate10/` | the classic, multi-grouping and all workloads (r = m = 1) re-planned at 10 samples/s per series, for the end-to-end run (on the grid before sketch-bench#213's denser one) | current (e2e input) |
| `results/case-studies-6.3.1/` | the paper's §6.3.1 case studies (performance monitoring, anomaly detection): tables, every method's plan, and what each deploys | current |
| `results/case-studies/` | the paper's §6.3.1 case studies (performance monitoring, anomaly detection), each swept one knob at a time: where ASAP's plan changes, and what every method deploys on the base workloads | current |
| `results/autosketch-vs-asap-inputs/` | the accuracy curves every run reads (sketch-bench#138's study), and sketch-bench#138's tables and raw records | current (shared input) |
| `results/autosketch-vs-asap-synthetic/`, `results/autosketch-vs-asap-synthetic-sla/` | the 2026-10-08 results (sketch-bench#138): before roll-ups and the KLL footprint fix | superseded by `-hydra` |
| `data/autosketch-eval/table.json` | the trace workloads (Alibaba 2022, Google 2011, BOOM) | not yet rerun on cost by use |
Expand Down
4 changes: 2 additions & 2 deletions rqe-optimizer/docs/e2e_planning_flow.md
Original file line number Diff line number Diff line change
Expand Up @@ -189,8 +189,8 @@ machine to itself.
It refuses a point that appears in two parts and mismatched headers.
4. Run `--phase optimizer-cost` alone on one machine.

On asap_sketchlib 0.3.0, five 56-core nodes run the default grid and the
targeted 1e9 points in about 10 minutes.
Five 56-core nodes run the default grid and the targeted 1e9 points in
about 10 minutes.

`SaturationCurves::load` expects exactly this layout:

Expand Down
9 changes: 5 additions & 4 deletions rqe-optimizer/docs/rqe_optimizer_hydra.md
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@ schema (the hyperperiod there is `H`).

## 1. What Hydra is in sketch-bench

All five combinations wrap `asap_sketchlib::Hydra` (0.3.0; Manousis et al.,
All five combinations wrap `asap_sketchlib::Hydra` (Manousis et al.,
VLDB 2022, arXiv:2208.04927). The `polars` rows of the same names are exact
baselines.

Expand Down Expand Up @@ -46,7 +46,8 @@ baselines.
| `hydra-univmon` | UnivMon: heap 1000, 5 × 2048, 8 layers; grid 2 × 8 | `R·W·layers·((sr·sc + sr)·8 + heap·entry)` |

`hydra-univmon` is registered as four statistics: cardinality, L1, L2 and
entropy. Each footprint adds `(R·W + 1)` cell headers.
entropy. The reported footprint is `Hydra::memory_bytes`: every cell's own
`memory_bytes`, plus the grid struct and the schema's labels.

## 2. Accuracy guarantee per combination

Expand Down Expand Up @@ -234,12 +235,12 @@ them all; `hydra-kll` is measured in that same run, alongside `hydra-cms`,
`hydra-cs`, `hydra-hll` and `hydra-univmon`, under every item below.

1. **Target grouping.** Comparators score only column 0
(`sketch-bench/aqpbm-cli/src/rows/mod.rs`); they must score any named subset of the
(`sketch-bench/sketch-bench/src/rows/mod.rs`); they must score any named subset of the
schema, and the wrapper must query any subset, not only prefixes
(`hydra_shared.rs` `query`).
2. **Exact baseline.** The `polars` baselines build untagged subkeys
(`polars_shared.rs`), so overlapping label values alias there but not in
the 0.3.0 sketch (sketch-bench#74); tag them the same way.
the sketch (sketch-bench#74); tag them the same way.
3. **Data.** Hierarchical labels with set cardinality per label, per-parent
fan-out and skew; the result records `|Λ|`, the target grouping's
cardinality, every group's `N_q`, and the fanned-out mass `M`.
Expand Down
74 changes: 0 additions & 74 deletions rqe-optimizer/results/case-studies-6.3.1/README.md

This file was deleted.

Loading
Loading