dualpath-storage-bandwidth-optimization

Optimize disaggregated prefill-decoding LLM serving for multi-turn (agentic) workloads by introducing dual-path KV-cache loading. Traditional approach loads all KV-cache to prefill engines, saturating their storage network. DualPath loads to decoding engines first, then transfers via compute network (lower contention). Adaptive routing selects path based on…

Independent GitHub snapshot · Automated classification · Original documentation governs