Per analysis/unified_routing_fix_review.md #2, several docs still presented the retired single-argmin + PUSH-migration design as the final algorithm. Mark them superseded and document the current hybrid direction (commit255c8e6). - REPORT.md §1.1 / §3.9: add errata callout and section header noting the "Final Design" framing was retired aftercc6e562/ 4c583f2; point readers to docs/migration-policy-design.md. - docs/migration-policy-design.md: rewrite. Opens with the current hybrid algorithm (LMetric base + cache_ratio>0.5 affinity gate + tie-breaker), then a "What Was Retired" commit table, then the old Approach A numbers preserved as "Historical Baseline-Mode Comparison". - analysis/research_findings.md §2.2 / §5: correct the LMetric framing. LMetric isn't "neutralized by affinity constraints" (pure --policy lmetric has no affinity at all); it converges to similar placements because P_tokens includes new_uncached_tokens, giving it implicit soft affinity. - analysis/elastic_hypotheses.md: same LMetric correction in the "DOESN'T work" summary, plus a footer cross-referencing the current routing direction. - analysis/unified_routing_fix_review.md: track this file (was untracked); it is the review handoff cited from the updated docs. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
13 KiB
Unified Routing Fix Review Handoff
Date: 2026-05-25
This is the corrected review handoff after reading the git history. The key change from the previous draft is that we should not restore the old single-argmin / PUSH-migration design. That path was implemented, measured, and then discarded or simplified by later commits.
Executive Summary
The latest commit history says the current direction is:
- Use baseline mode / no Mooncake for the winning comparison against LMetric.
- Use LMetric-style load balancing as the base.
- Add explicit affinity only for sessions with high accumulated cache.
- Do not re-enable PD-sep offload or session migration unless the transfer mechanism is fundamentally reworked.
The main fixes now are cleanup, documentation consistency, tests, and reproducibility. The biggest risk is that stale docs still describe abandoned schemes as "final design".
Evidence From Git History
Relevant commits:
| Commit | Meaning | Outcome |
|---|---|---|
6b255fa |
Implemented single argmin(expected_latency) over local / PUSH / cold paths |
Later superseded |
5892739 |
Added soft affinity because pure argmin overloaded cache-source instances | Pure argmin was unstable |
2b9eae0 |
Reported Unified v3 with 116 PUSH migrations | TTFT improved, but TPOT/E2E tail tradeoff existed |
cdf8349 |
Added real cache sync and cached-prefill-on-C architecture | Fixed false PUSH and direct-read issues |
4b50c5a |
Fixed queue model and hard overload gate | Reduced imbalance, but still on offload path |
e991960 / 5772149 |
Tried forced session migration triggers | Reverted |
cc6e562 |
Reverted Approach B migration | 57 migrations made HEAVY TTFT p90 regress 15.9s -> 59.1s |
bf4469a |
Tried more accurate push cost / gate alignment | Later reverted |
4c583f2 |
Reverted relaxed gate / push-cost fix | 134 offloads made E2E p90 37s -> 82s |
448361c |
Updated design doc: baseline no-Mooncake Unified beats LMetric | PD-sep offload degrades |
255c8e6 |
Replaced full cost model with hybrid routing | Current direction: LMetric LB + high-cache affinity |
Do not ask the implementer to "restore real three-way argmin". That was the wrong instruction in the previous draft.
Current Intended Algorithm
From 255c8e6, the current algorithm should be documented as:
if session has an affinity instance:
if cache_ratio_on_affinity > 0.5
and affinity_instance.num_requests <= avg_num_requests * overload_factor:
route to affinity instance
else:
route by LMetric
else:
route by LMetric
LMetric remains:
score = (pending_prefill_tokens + new_uncached_tokens) * num_requests
new_uncached_tokens = input_length - estimated_cache_hit
This is no longer the old expected-latency migration model.
P0 Fixes
1. Remove Stale PUSH-Migration Code From the Current Unified Branch
Location: scripts/cache_aware_proxy.py, _handle_combined.
Problem:
After 255c8e6, unified is a hybrid policy, but the function still contains
an unreachable block guarded by best_needs_push = False. That block references
variables from the removed cost model such as best_cache_idx,
best_cache_hit, and _current_offloads.
Fix:
- In the
unifiedbranch, delete the unreachableif best_needs_push:block. - Keep helper functions like
_handle_cached_prefill_offloadonly if another live mode still calls them. - Update the
_handle_combineddocstring so it no longer says Unified always computesqueue + prefill + transfer.
Why:
The code is currently safe only because the branch is unreachable. Leaving it there makes future changes dangerous and makes reviewers think migration is still part of the active policy.
Verification:
rg "best_needs_push|best_cache_idx|push_cache_hit" scripts/cache_aware_proxy.pyshould show no stale references inside the activeunifiedpath.pytest -q.
2. Reconcile Documentation With the Latest Commits
Locations:
REPORT.mddocs/migration-policy-design.mdanalysis/research_findings.mdanalysis/elastic_hypotheses.md
Problem:
Several docs still present old or contradictory conclusions:
REPORT.mdsection 3.9 still calls single argmin / PUSH migration the "Final Design".docs/migration-policy-design.mddescribes a baseline-mode additive cost model, while HEAD implements hybrid LMetric + high-cache affinity.analysis/research_findings.mdsays LMetric is neutralized by session affinity constraints, but later corrected LMetric results say cache-hit scoring creates implicit soft affinity.
Fix:
- Add an explicit "Superseded by git history" note near
REPORT.mdsection 3.9:6b255fa/5892739/2b9eae0were explored.cc6e562rejected forced migration.4c583f2rejected relaxed offload / high-offload configurations.255c8e6is the current implementation direction.
- Update
docs/migration-policy-design.mdto either:- describe the current hybrid algorithm, or
- clearly state that the additive cost model is the previous Approach A and not the latest code.
- Mark old analysis sections as superseded rather than deleting them.
Why:
The docs caused the previous bad review recommendation. Future reviewers need to know which ideas were already tested and rejected.
Verification:
rg "Final Design|argmin\\(expected_latency\\)|PUSH_MIGRATE|Approach B" REPORT.md docs analysisshould show clear superseded labels where appropriate.
3. Preserve the LMetric Baseline Separately From Unified Hybrid
Location: scripts/cache_aware_proxy.py.
Problem:
Pure --policy lmetric is the baseline being compared against. Unified hybrid
uses LMetric internally but should not accidentally change the LMetric baseline
behavior.
Fix:
- Keep
pick_instance_lmetricas the pure corrected LMetric implementation. - Put any hybrid-specific tie-breakers or affinity logic outside
pick_instance_lmetric, under--policy unified. - In breakdown logs, record whether a Unified request used
affinityorlmetric_fallback.
Why:
If the baseline and Unified share hidden behavior, future A/B comparisons become invalid.
Verification:
- Unit test:
--policy lmetricnever uses session affinity. - Unit test:
--policy unifiedcan use affinity only when cache ratio and load gates pass.
P1 Fixes
4. Fix the Unified Hybrid / LMetric Fallback Empty-Batch Degeneracy
Problem:
LMetric score is P_tokens * BS. When BS = num_requests = 0, every instance
gets score 0, so tie-break chooses instance 0. docs/migration-policy-design.md
explicitly lists avoiding this as one reason Unified beats LMetric.
The latest hybrid falls back to LMetric, so it may reintroduce this issue for new sessions when all instances are idle.
Fix:
- Do not change the pure
--policy lmetricbaseline. - For
--policy unifiedfallback only, add a deterministic secondary key:- primary:
P_tokens * BS - secondary when scores tie:
new_uncached_tokens, thennum_requests, then a round-robin or least-recently-used instance index.
- primary:
- Record tie-break count in breakdown or stats.
Why:
This preserves fair LMetric comparison while preventing Unified hybrid from degenerating to instance 0 under empty or near-empty load.
Verification:
- Unit test with all
num_requests = 0: Unified should not always choose index 0 across repeated new sessions. - Confirm pure LMetric test still matches the OSDI-style baseline.
5. Add Tests for the Current Hybrid Policy
Problem:
Existing tests cover older pick_instance and pure LMetric behavior, but not
the current unified branch introduced by 255c8e6.
Fix:
Add tests for:
- High-cache session sticks to affinity instance when not overloaded.
- High-cache session breaks affinity when
num_requestsexceeds the overload gate. - Low-cache session falls back to LMetric.
- New session falls back to LMetric with the Unified-specific tie-breaker.
- Breakdown policy is recorded as
affinityorlmetric_fallback.
Why:
This prevents future drift back toward the discarded migration cost model or accidental changes to the LMetric baseline.
Verification:
pytest -q.- Tests should run without live vLLM.
6. Treat overload_factor Changes as Experiments, Not Silent Fixes
Observation:
The current worktree has an uncommitted change:
Settings.overload_factor: 2.0 -> 1.5
But the earlier H7 sweep found overload-factor tuning largely ineffective /
within noise. This is recorded in analysis/elastic_hypotheses.md.
Fix:
- Do not silently commit a default change to
1.5without a paired benchmark. - If testing
1.5, make it an experiment tag/config value, not a new default. - Keep docs and CLI defaults synchronized.
Why:
The prior experiment says imbalance was mostly workload/session skew, not the threshold. A default change without evidence will create another reproducibility gap.
Verification:
- Run paired
2.0vs1.5after the hybrid tests exist. - Report E2E p50/p90, TTFT p90, APC distribution, and GPU util imbalance.
7. Standardize Breakdown Fields for Hybrid Routing
Problem:
The current breakdown logs do not clearly expose why Unified chose affinity vs LMetric fallback.
Fix:
For each request under --policy unified, log:
policy: "unified"decision: "affinity" | "lmetric_fallback"affinity_idxchosen_idxaffinity_cache_hitaffinity_cache_ratioaffinity_num_requestsavg_num_requestsfallback_scoretie_break_used
Why:
The latest performance difference versus LMetric is small. Without decision logs, it is hard to tell whether Unified is actually exercising the intended high-cache affinity behavior.
Verification:
breakdown.jsoncan answer: how many requests used affinity, how many used fallback, and what latency/APC each group saw.
P2 Experiment Fixes
8. Re-run Paired LMetric vs Unified Hybrid Benchmarks
Problem:
docs/migration-policy-design.md says the 2% mean improvement needs multi-run
verification. Local raw outputs for the May 25 final comparison are not present
in this workspace.
Fix:
- Run 3-5 fresh paired trials.
- Same trace, same vLLM build, same machine, same restart procedure.
- Compare:
- pure LMetric
- current Unified hybrid
- optionally Linear/session-sticky as a reference
Metrics:
- TTFT mean/p50/p90/p99
- TPOT mean/p50/p90/p99
- E2E mean/p50/p90/p99
- errors/timeouts
- aggregate APC
- per-instance APC distribution
- per-instance request count and token count
- GPU util mean/std/imbalance
Why:
The current reported win is small enough that run-to-run noise matters.
Verification:
- Commit or save artifacts under a date/tagged output directory:
config.json,metrics.jsonl,metrics.summary.json,breakdown.json,gpu_util.csv, final vLLM APC snapshot, git commit hash.
9. Do Not Re-open PD-Sep Offload Without a New Transfer Mechanism
Rejected paths:
- Full PD separation: decode KV memory wall.
- Elastic P2P RDMA offload: transfer and scheduling overhead exceed benefit.
- Cache-gate offload: improves balance for colocated survivors but offloaded requests pay RDMA penalty.
- Approach B session migration: 57 migrations made HEAVY TTFT p90 much worse.
- Relaxed gate / many offloads: 134 offloads made E2E p90 much worse.
Future work can revisit migration only if one of these changes first:
- layerwise / pipelined KV transfer
- multi-machine P role so P work does not compete with D on the same GPU pool
- exact vLLM state / cache residency exposed to the router
- production-concurrency benchmark showing decode SLO pressure large enough to amortize transfer overhead
Why:
The same local mechanism has already failed multiple times. Repeating it with another threshold is unlikely to help.
10. Restore Production-Concurrency Evaluation as a Separate Track
Problem:
Several conclusions were made at 1-2 req/GPU, while production is estimated at 8-15 req/GPU. Higher concurrency is where prefill-decode interference appears.
Fix:
- Restore or replace
--max-inflight-sessionsfor controlled concurrency. - Run at 64 and 128 active sessions.
- Treat this as a new experiment track, not as a reason to resurrect old migration code immediately.
Why:
At high concurrency, the design pressure may change. But the implementation should first prove the current hybrid baseline cleanly.
Suggested Implementation Order
- Update docs to mark discarded migration/offload schemes as superseded.
- Remove stale unreachable PUSH code from the current Unified branch.
- Add tests for current Unified hybrid behavior.
- Add a Unified-only tie-breaker for LMetric fallback empty-batch cases.
- Add breakdown fields for hybrid routing decisions.
- Decide whether the local
overload_factor=1.5diff is an experiment or should be dropped. - Run paired multi-run LMetric vs Unified hybrid benchmarks.
- Only after that, open a separate high-concurrency experiment track.
Review Checklist
- Does the code still mention single
argmin(expected_latency)as current behavior? If yes, update it or mark it superseded. - Is any migration/offload code reachable under
--policy unified? If yes, require a new experiment plan because recent history rejects it. - Does pure
--policy lmetricremain pure and affinity-free? - Does
--policy unifiedclearly log when it used affinity vs LMetric fallback? - Are default parameter changes backed by paired results?
- Can another reviewer reproduce the result from committed scripts and saved artifacts?