发表机构
Tsinghua University; Meituan（清华大学; 美团）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
AFD-Ledger是离线分析配置系统，可优化AFD与同置部署的硬件分配，减少部署评估量，预测吞吐量偏差小，揭示同构/异构AFD及角色硬件改进的部署规律。
AI 中文摘要
注意力-前馈网络（FFN）拆分（AFD）正成为服务混合专家（MoE）语言模型的有前景架构。现有AFD系统虽提升了拆分执行的效率，但留下一个部署问题未解答：在相同模型、工作负载、每输出令牌时间（TPOT）服务水平目标（SLO）、硬件预算、硬件目录及运行时能力下，AFD是否比最佳同置部署提供更高吞吐量？解答该问题需联合优化两种架构的硬件分配与部署组织，使穷尽配置成本过高。本文提出AFD-Ledger，这是一种离线分析配置系统，采用分析执行模型与评估受限的硬件搜索，分别为AFD与同置部署进行配置。在穷尽配置可行的部署空间中，AFD-Ledger减少68.8%至83.5%的完整部署评估，同时仍恢复全局最优部署。在三个实际LongCat 2.0部署上，它保留正确的架构决策，同时预测AFD与同置部署的吞吐量，与测量值的偏差在6.6%至9.6%之间。利用此验证框架，本文表明：同构AFD仅在少数研究设置中提升固定预算吞吐量；异构AFD需要部署级硬件互补性，而非启发式设备选择；特定角色的硬件改进主要在其通过跨越部署能力-价格边界实现更好部署组织时才重要。
英文摘要
Attention--Feed-Forward Network (FFN) Disaggregation (AFD) is emerging as a promising architecture for serving Mixture-of-Experts (MoE) language models. While existing AFD systems improve the efficiency of disaggregated execution, they leave a deployment question unanswered: under the same model, workload, time-per-output-token (TPOT) service-level objective (SLO), hardware budget, hardware catalog, and runtime capabilities, does AFD provide higher throughput than the best collocated deployment? Answering this question requires jointly optimizing hardware assignment and deployment organization for both architectures, making exhaustive provisioning prohibitively expensive. We present AFD-Ledger, an offline analytical provisioning system that independently provisions AFD and collocated deployments using an analytical execution model and an evaluation-bounded hardware search. Across deployment spaces where exhaustive provisioning is feasible, AFD-Ledger reduces complete deployment evaluations by 68.8%--83.5% while still recovering the globally optimal deployment. On three physical LongCat 2.0 deployments, it preserves the correct architecture decision while predicting AFD-to-collocated throughput within 6.6%--9.6% of measurement. Using this validated framework, we show that homogeneous AFD improves fixed-budget throughput in only a minority of the studied settings, heterogeneous AFD requires deployment-level hardware complementarity rather than heuristic device selection, and role-specific hardware improvements matter primarily when they enable better deployment organizations by crossing deployment capability--price boundaries.
Comments14 pages, 14 figures, 2 tables
