Consistent-hash ring with weighted virtual nodes and zone-aware replica placement
Exact prompt
# Consistent-hash ring with weights and zone-aware replicas `ring.py` places cache keys on storage nodes. The draft uses the wrong hash, ignores weights and zones, and is slow. Rewrite it (Python 3 stdlib only) so that `Ring(nodes, vnodes=64)`, `Ring.owner(key)`, `Ring.replicas(key, n)` and `moves(old_ring, new_ring, keys)` behave exactly as below. ## Hashing 1. `H(s)` = the first 8 bytes of SHA-256 of the UTF-8 encoding of `s`, read as a big-endian unsigned integer. 2. `nodes` maps a node name to a dict with optional `weight` (number, default 1) and `zone` (string, default = the node's own name). `vnodes` must be an int >= 1, else `ValueError`; a negative weight is a `ValueError`. 3. A node gets `k = floor(vnodes * weight + 0.5)` virtual points, but at least 1 if weight > 0. A node with weight 0 gets none and does not exist for the ring. Point i (0 <= i < k) of node `name` is at position `H(f"{name}#{i}")`. 4. The ring is all points sorted by (position, node name, i). ## Lookups 5. `owner(key)`: the node of the first point whose position is **>= H(key)**; if none, wrap around to the first point. An empty ring raises `LookupError`. 6. `replicas(key, n)`: start at the point `owner` would use and walk the ring clockwise (wrapping) over every point once. Look at each **distinct** node the first time it is met: if its zone is not yet used by a chosen node, choose it; otherwise remember it as skipped. Stop as soon as n nodes are chosen. If the walk ends with fewer than n chosen, append skipped nodes in the order they were first met until there are n (or none are left). `n < 1` gives `[]`. The owner is always first. 7. `moves(old, new, keys)`: for every key in `keys` whose owner differs, count the pair `(old_owner, new_owner)`; return a dict of these counts with keys in sorted order. 8. **Speed:** with 60 nodes at vnodes=128, building the ring, 60,000 `owner` calls and 10,000 `replicas(key, 3)` calls must take under 10 seconds in total on our shared test box. ## Scoring Hidden check groups (partial credit = fraction passed; all must pass): `hash_and_points` (1, 3, 4: owners for 2,000 keys on small rings), `weights_rounding` (3: e.g. weight 0.5/0.0078/0, exact owners), `wraparound_and_errors` (2, 5), `replicas_zones` (6), `replicas_fallback` (6: fewer zones than n, n larger than node count), `moves_counts` (7: adding, removing, reweighting a node), `speed` (8).
Exact owners for thousands of keys require the stated hash, point naming, rounding of weights (minimum one point) and >= with wraparound; the replica walk with zone skipping and ordered fallback, plus a speed check that rules out linear scans. Original benchmark task written for usecaserank.com (not based on a specific public post). Headless, offline run on a reproducible fixture. Each run's notes say how its cost was measured (estimates are marked).
Success criteria: 1. hash_and_points 2. weights_rounding 3. wraparound_and_errors 4. replicas_zones 5. replicas_fallback 6. moves_counts 7. speed (< 10 s)
Best model for this task so far: Claude Haiku 5.5 (effort default (no --effort flag) · thinking on) (success, $0.0081, 52m).
All runs (4)
100% pass rate across 4 eligible runs: 4 pass, 0 partial, 0 fail. Partial and failed runs are listed below with everything else.ⓘ
| Model | Cost | Time | Step-insⓘ | Autonomy | Outcome | Trustⓘ | Date | Compare |
|---|---|---|---|---|---|---|---|---|
| OGPT-6.1 Soldefault · thinking on | $0.07 | 1m | 0 | Fully autonomous | pass | Trust 80% | 2026-10-22 | |
| AClaude Haiku 5.5default · thinking on | $0.0081 | 52m | 0 | Fully autonomous | pass | Trust 75% | 2026-10-20 | |
| AClaude Opus 5.5default · thinking on | $0.17 | 34s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-20 | |
| AClaude Sonnet 5.5default · thinking on | $0.07 | 19s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-20 |
Embed this result
A live badge for this use case. It links back here and updates as new runs are published.
Get new runs and ranking changes weekly
Which models got better or cheaper on real tasks like this one.