SeaLLM shares GPU resources across multiple LLM services with a preemptive, service-characteristic-aware scheduler, search-based placement, adaptive replacement, and a merged-block unified KV cache, cutting normalized latency by up to 13.6x.
https://grpc.io, 2021
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SeaLLM: Service-Aware and Latency-Optimized Resource Sharing for Large Language Model Inference
SeaLLM shares GPU resources across multiple LLM services with a preemptive, service-characteristic-aware scheduler, search-based placement, adaptive replacement, and a merged-block unified KV cache, cutting normalized latency by up to 13.6x.