MoE 模型推理的瓶颈终于被系统性地拆解了——Attention-FFN 解耦让吞吐量提升到传统方案无法企及的水平，做大规模 MoE 推理部署的团队可以直接参考论文中的分区策略来优化集群。
这篇论文系统研究了混合专家（MoE）大模型推理中的多级解耦策略，从分块预填充聚合、预填充-解码解耦到最新的算子级 Attention-FFN 解耦（AFD）。AFD 将注意力计算和 MoE-FFN 执行分别部署在不同 GPU 组上，以应对 MoE 模型中内存受限的注意力、计算密集的专家 FFN 以及 MoE 路由通信带来的异构资源需求。通过结合设备级内核测量和高保真网络模拟的框架，论文在真实工作负载下评估了各级解耦的收益与局限。结果表明，在严格的 TTFT/TPOT 服务等级目标下，AFD 在 DeepSeek-V3.2 上可维持约 4000 tokens/s 的系统吞吐量，而传统非 AFD 部署无法满足要求。论文还给出了根据工作负载和模型架构划分注意力与 FFN 的实用建议，为当前机架/集群级部署及未来解耦 AI 基础设施提供了设计原则。
How Far Can Disaggregation Go? A Design-Space Exploration of Attention-FFN Disaggregation for Efficient MoE LLM Serving
Modern large language model (LLM) inference has progressively disaggregated to keep pace with growing model sizes and tight TTFT and TPOT service-level objectives: from chunked-prefill aggregation, to prefill-decode (P/D) disaggregation, and most recently to operator-level Attention-FFN Disaggregation (AFD). This trend is especially important for mixture-of-experts (MoE) models, where memory-bound attention, compute-intensive expert FFNs, and MoE dispatch/combine communication create distinct resource demands. AFD further exposes this heterogeneity by placing attention and MoE-FFN execution on separate GPU groups. Each level of disaggregation deepens the scheduling design space across workload characteristics, resource allocation, and interconnect topology, raising the central question: when does each level actually pay off? We systematically characterize this trade-off for MoE inference across realistic workloads spanning input/output sequence lengths, prefix-KV reuse, and per-user latency constraints. Using chunked-prefill and P/D disaggregation as baselines, we study the benefits and limits of AFD at scale through a framework that fuses on-device kernel measurements with high-fidelity network simulation. Under strict TTFT/TPOT SLOs, AFD sustains around 4k tokens/s of system throughput on DeepSeek-V3.2 across chat, coding, and agentic-coding workloads, where non-AFD deployments are infeasible. We distill concrete takeaways for jointly optimizing throughput and interactivity, including how to partition attention and FFN across GPUs as a function of workload and model architecture, providing design principles for current rack- and cluster-scale deployments as well as future disaggregated AI infrastructure.