Claude Code Agent dc4867c270 feat(snapshot): D→P RDMA link Phase 1 — minimal byte transport ...

A thin wrapper around mooncake.engine.TransferEngine that does
one-sided RDMA writes between two SnapshotPeer endpoints. Bypasses
SGLang's MooncakeKVManager (which is hard-gated to PREFILL/DECODE
roles via add_transfer_request assertion at conn.py:1563) so the
D→P direction doesn't require invasive role-axis changes upstream.

Smoke test (two subprocess.Popen processes, mlx5_60, 127.0.0.1):
  1 KB    9.0 ms   (one-time openSegment handshake)
  16 KB   0.04 ms  3.5 Gbps
  1 MB    0.10 ms  82 Gbps
  16 MB   0.58 ms  232 Gbps
  64 MB   1.70 ms  316 Gbps   (~80% of NDR 400G line rate)

All 5 sizes pass SHA256 verification end-to-end.

Files:
  src/agentic_pd_hybrid/snapshot_link.py — SnapshotPeer, SnapshotEndpoint
  scripts/snapshot_link_receiver.py      — child-process receiver
  scripts/smoke_snapshot_link.py         — sender + verifier
  docs/D_TO_P_PHASE1_LINK_ZH.md          — phase 1 acceptance doc

Next: Phase 2 (D-side scheduler commit hook), Phase 3 (P-side prefill
bypass with snapshot KV). See docs/D_TO_P_SYNC_DESIGN_ZH.md §5.

2026-05-13 00:55:55 +08:00

..

analysis

fix(figures): GPU utilization figure annotation/headroom polish

2026-05-11 22:28:39 +08:00

convert_audit_to_trace.py

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

convert_inferact_to_trace.py

feat(scripts): cu12.8 env wrapper + Inferact trace converter

2026-05-12 00:10:06 +08:00

run_all_experiments.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

run_exp_a_pd_disagg.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

run_exp_b1_dp_colo_rr.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

run_exp_b2_dp_colo_cache_aware.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

run_exp_b_pd_colo.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

run_exp_c_kvcache_centric.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

sample_trace_subset.py

feat(experiments): E1 sweep on 50-session deterministic subset

2026-05-12 00:21:36 +08:00

setup_env.sh

feat(env): MC_TRANSFER_TIMEOUT=1800s default in setup_env + stack

2026-05-12 11:45:09 +08:00

smoke_snapshot_link.py

feat(snapshot): D→P RDMA link Phase 1 — minimal byte transport

2026-05-13 00:55:55 +08:00

smoke_test.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

snapshot_link_receiver.py

feat(snapshot): D→P RDMA link Phase 1 — minimal byte transport

2026-05-13 00:55:55 +08:00

sweep_backpressure_smoke.sh

feat(kvc): add backpressure smoke sweep + analyzer (and v6 p1 profile script)

2026-05-06 21:29:56 +08:00

sweep_e1_naive_1p3d.sh

feat(experiments): E1 sweep on 50-session deterministic subset

2026-05-12 00:21:36 +08:00

sweep_e2_kvc_v2_rdma.sh

feat(experiments): E2 sweep — KVC v2 + RDMA on the matched subset

2026-05-12 00:49:53 +08:00

sweep_e3_kvc_v2_loadfloor_rdma.sh

feat(experiments): E3 sweep — KVC v2 + RDMA + load-floor bonus

2026-05-12 11:45:09 +08:00

sweep_kvc_qwen3_30b.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

sweep_tp1_configs.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

sweep_tp1_v2_fixed.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

sweep_tp1_v3_kvaware.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

sweep_tp1_v4_cap16.sh

docs: KVC v1-v4 debug journey + raise session soft_cap to 16

2026-04-28 21:10:41 +08:00

sweep_tp1_v5_baseline_rerun_exp2.sh

profile(kvc): rewrite v5+profile report after critic audit + P0/P1 instrument

2026-04-29 22:29:21 +08:00

sweep_tp1_v5_optD_profile.sh

profile(kvc): add D KV pool timeseries poller + analyzer for v6 root-cause

2026-04-29 20:04:21 +08:00

sweep_tp1_v5_optD.sh

feat(kvc): Option D - delegate seed/reseed admission to D worker

2026-04-28 23:40:03 +08:00

sweep_tp1_v6_p1_profile.sh

feat(kvc): add backpressure smoke sweep + analyzer (and v6 p1 profile script)

2026-05-06 21:29:56 +08:00

sweep_ts1_kvc_n3_plus_dp.sh

feat(kvc): session migration with reset-on-success + direct-append threshold tuning

2026-05-09 01:18:13 +08:00

sweep_ts1_migration_v1.sh

feat(kvc): session migration with reset-on-success + direct-append threshold tuning

2026-05-09 01:18:13 +08:00

sweep_ts1_migration_v2.sh

feat(kvc): session migration with reset-on-success + direct-append threshold tuning

2026-05-09 01:18:13 +08:00