Draft-li-kvcache-efficient-storage-00
Comments
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Summary
KVSP is an HTTP-based Internet-Draft from Li et al. (IETF, November 2026) specifying how separate components of LLM serving systems share attention key-value cache blocks. It defines compatibility descriptors, keyed block identifiers, replica indexing with per-layer residency, leased prefetch, session checkpoints for migration, and resharding for sequence-parallel caches.