MoE-Prism：通过模型-系统协同设计解耦整体式专家以实现弹性MoE服务
MoE-Prism: Disentangling Monolithic Experts for Elastic MoE Services via Model-System Co-Designs
AI总结：
MoE-Prism通过模型-系统协同设计将MoE的整体专家分解为细粒度子专家，实现请求级计算弹性，扩展路由工作点数量，提升吞吐量并降低首包时间，验证了实用弹性MoE服务的可行性。
摘要：
混合专家（Mixture-of-Experts，MoE）通过稀疏激活扩展模型容量，正成为大语言模型（LLMs）的重要架构。然而，现有MoE服务系统通常在固定路由配置下执行所有请求，限制了其利用不同请求间异构计算需求的能力。路由top-k决定了每个token激活的路由专家数量，直接控制路由专家的计算量，为请求级计算弹性提供了自然机制。但要实现该能力，需要更细粒度的路由单元和针对异构路由预算的高效运行时执行。我们提出MoE-Prism，一种用于MoE服务中请求级计算弹性的模型与系统支持框架。MoE-Prism将整体式专家分解为细粒度子专家，以暴露更密集的路由工作点，并提供支持k的服务运行时，可在面向吞吐量和延迟敏感的工作负载下有效服务异构路由预算。我们在vLLM基础上实现MoE-Prism，并在三个代表性MoE模型上进行评估。MoE-Prism将可用路由工作点数量扩展了4倍，将离线推理吞吐量提升高达33.9%，并降低了异构工作负载下的在线服务首包时间（TTFT）。这些结果证明了具备请求级路由目标的实用弹性MoE服务的可行性。
英文摘要：
Mixture-of-Experts (MoE) scales model capacity through sparse activation, and is becoming an important architecture for large language models (LLMs). However, existing MoE serving systems typically execute all requests under a fixed routing configuration, limiting their ability to exploit heterogeneous computation requirements across requests. Routing top-$k$, which determines the number of routed experts activated per token, directly controls routed-expert computation and provides a natural mechanism for request-level compute elasticity. Realizing this capability, however, requires finer-grained routing units and efficient runtime execution for heterogeneous routing budgets. We present \textsc{MoE-Prism}, a model and system support framework for request-level compute elasticity in MoE serving. \textsc{MoE-Prism}decomposes monolithic experts into fine-grained sub-experts to expose denser routing operating points and provides a $k$-aware serving runtime that effectively serves heterogeneous routing budgets under both throughput-oriented and latency-sensitive workloads. We implement \textsc{MoE-Prism} on top of vLLM and evaluate it on three representative MoE models. \textsc{MoE-Prism} expands the number of available routing operating points by $4\times$, improves offline inference throughput by up to 33.9\%, and reduces online serving TTFT under heterogeneous workloads. These results demonstrate practical elastic MoE serving with request-level routing targets.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
