TENT: A Declarative Slice Spraying Engine for Performant and Resilient Data Movement in Disaggregated LLM Serving
1 directory member surfaced this signal.
“TENT reshapes language model serving with a new orchestration engine, improving data movement efficiency and resilience in GPU clusters. It achieves 1.36× higher throughput and 26% faster updates, adapting to interconnect variations and cutting operational …” evidence ↗
“TENT reshapes language model serving with a new orchestration engine, improving data movement efficiency and resilience in GPU clusters. It achieves 1.36× higher throughput and 26% faster updates, adapting to interconnect variations and cutting operational …”