mlsanityData to 14 Oct 2026

AFD-Ledger: Deployment Provisioning for Attention--FFN Disaggregation

arXiv 2608.04502 ↗First posted 5 Aug 2026cs.DC cs.AIcited by 0

On the map: Mixture-of-experts → Efficient MoE training and inference systems

Abstract (arXiv)

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.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Efficient MoE training and inference systems, one of 12 directions in Mixture-of-experts. Builds on 17 of its 27 references that are in the map. First posted on arXiv 5 Aug 2026.

Builds on

17 of 27 references.

Closest ancestors · 7

Same direction, newest first

  1. 2026How Far Can Disaggregation Go? A Design-Space Exploration of Attention-FFN Disaggregation for Efficient MoE LLM Serving27 May
  2. 2026Revealing the Challenges of Attention-FFN Disaggregation for Modern MoE Models and Hardware Systems10 Feb
  3. 2025Efficient MoE Inference with Fine-Grained Scheduling of Disaggregated Expert Parallelism25 Dec
  4. 2025Expert-as-a-Service: Towards Efficient, Scalable, and Robust Large-scale MoE Serving22 Sep
  5. 2022Tutel: Adaptive Mixture-of-Experts at Scale7 Jun
  6. 2022DeepSpeed-MoE: Advancing Mixture-of-Experts Inference and Training to Power Next-Generation AI Scale14 Jan
  7. 2021FastMoE: A Fast Mixture-of-Expert Training System24 Mar

Other references · 10

  1. 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsitycited by 3,073
  2. 2020GShard: Scaling Giant Models with Conditional Computation and Automatic Shardingcited by 1,631
  3. 2023Splitwise: Efficient generative LLM inference using phase splittingcited by 558
  4. 2024Taming Throughput-Latency Tradeoff in LLM Inference with Sarathi-Servecited by 526
  5. 2024Mooncake: A KVCache-centric Disaggregated Architecture for LLM Servingcited by 196
  6. 2023SARATHI: Efficient LLM Inference by Piggybacking Decodes with Chunked Prefillscited by 193
  7. 2024Inference without Interference: Disaggregate LLM Inference for Mixed Downstream Workloadscited by 121
  8. 2024MemServe: Context Caching for Disaggregated LLM Serving with Elastic Memory Poolcited by 68
  9. 2026KVServe: Service-Aware KV Cache Compression for Communication-Efficient Disaggregated LLM Servingcited by 7
  10. 2026Frontier: Towards Comprehensive and Accurate LLM Inference Simulationcited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Analytical Provisioning for Attention-FFN Disaggregated LLM Serving under Stochastic Workloads
  2. 2025FedQuad: Adaptive Layer-wise LoRA Deployment and Activation Quantization for Federated Fine-Tuning
  3. 2026Energy-Efficient LLM Serving via Disaggregated Attention--FFN and Flexible Frequency Scaling
  4. 2025Injecting Adrenaline into LLM Serving: Boosting Resource Utilization and Throughput via Attention Disaggregation
  5. 2026Beyond Fluent Generation: A CPU Reliability Benchmark for MCP-Style Tool Calling in Sub-2B Small Language Models for Edge Deployment
  6. 2025BanaServe: Unified KV Cache and Dynamic Module Migration for Balancing Disaggregated LLM Serving in AI Infrastructure
  7. 2026Optimizing AI Inference Across the Deployment Stack
  8. 2026OpWeave: Flexible Operator Disaggregation for Heterogeneous LLM Serving