2 Commits

Author SHA1 Message Date
6a27f75337 Docs: reconcile routing docs with current hybrid direction
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 Mark them superseded and document the current hybrid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 (commit 255c8e6).

- REPORT.md §1.1 / §3.9: add errata callout and section header noting
  the "Final Design" framing was retired after cc6e562 / 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>
2026-05-25 10:47:14 +08:00
8e0c6e78b0 Add comprehensive research findings document
Synthesizes all experiments into a paper-ready analysis:
- Agentic workload characteristics vs chatbot/API
- Why PD-Sep, LMetric, elastic RDMA, chunk-size tuning don't work
- Why cache-aware session-sticky routing IS the key optimization
  (-60% TTFT, +24pp APC vs round-robin)
- System-level insights: prefill-decode interference threshold,
  Mooncake limitations, effective request weight after cache
- GPU balance → HEAVY TTFT -10.5% (demonstrated)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-05-23 07:16:31 +08:00