Gahow Wang 1262c9c22e Migration transfer-cost study: KV transfer is slow on busy GPUs
MIGRATION_TRANSFER_COST.md: under real load, migration KV transfer runs at
~3 GB/s vs ~10 GB/s idle. Decomposed (instruments + MB6 microbench) into
~55% RDMA-actual (HBM/PCIe contention with running kernels: 7.6->4.0 GB/s)
+ ~45% control-plane GIL starvation during long prefills. Reproduced on a
fresh upstream venv (byte-identical transfer path) -> upstream/hardware
inherent, not our patch. Layerwise is the wrong lever; the tax is structural
on a loaded agentic cluster. Includes mb6_transfer_under_load + run_mb6,
instrument_dst_migration/mooncake, and the dst/transfer decomposition analyzers.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-05-29 11:53:01 +08:00
Description
No description provided
48 MiB
Languages
Python 82.9%
Shell 17.1%