# Analytical Power-Aware Provisioning for Prefill-Decode Disaggregated AI Inference

arXiv: https://arxiv.org/abs/2609.24639 · HTML: https://mlsanity.com/paper/2609.24639/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Serving systems and request scheduling, one of 12 directions in Efficient LLM inference and serving.
- Builds on 30 of its 57 references that are in the map.
- First posted on arXiv 21 Sep 2026.

## Placed in

- Efficient LLM inference and serving / [Serving systems and request scheduling](https://mlsanity.com/inference/serving-systems-and-request-scheduling/)

## Builds on (most cited first, 30 of 30)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints](https://mlsanity.com/paper/2305.13245/) (2023)
- [SGLang: Efficient Execution of Structured Language Model Programs](https://mlsanity.com/paper/2312.07104/) (2023)
- [Splitwise: Efficient generative LLM inference using phase splitting](https://mlsanity.com/paper/2311.18677/) (2023)
- [Taming Throughput-Latency Tradeoff in LLM Inference with Sarathi-Serve](https://mlsanity.com/paper/2403.02310/) (2024)
- [The Carbon Footprint of Machine Learning Training Will Plateau, Then Shrink](https://mlsanity.com/paper/2204.05149/) (2022)
- [Vidur: A Large-Scale Simulation Framework For LLM Inference](https://mlsanity.com/paper/2405.05465/) (2024)
- [Offline Energy-Optimal LLM Serving: Workload-Based Energy Models for LLM Inference on Heterogeneous Systems](https://mlsanity.com/paper/2407.04014/) (2024)
- [TokenPowerBench: Benchmarking the Power Consumption of LLM Inference](https://mlsanity.com/paper/2512.03024/) (2025)
- [Optimizing LLM Inference: Fluid-Guided Online Scheduling with Memory Constraints](https://mlsanity.com/paper/2504.11320/) (2025)
- [TokenScale: Timely and Accurate Autoscaling for Disaggregated LLM Serving with Token Velocity](https://mlsanity.com/paper/2512.03416/) (2025)
- [Agentic Coding in the Wild: Characterizing GitHub Copilot Traces at Production Scale](https://mlsanity.com/paper/2608.00101/) (2026)
- [SweetSpot: An Analytical Model for Predicting Energy Efficiency of LLM Inference](https://mlsanity.com/paper/2602.05695/) (2026)
- [Arrow: Adaptive Scheduling Mechanisms for Disaggregated LLM Inference Architecture](https://mlsanity.com/paper/2505.11916/) (2025)
- [Taming the Chaos: Coordinated Autoscaling for Heterogeneous and Disaggregated LLM Inference](https://mlsanity.com/paper/2508.19559/) (2025)
- [DualScale: Energy-Efficient Disaggregated LLM Serving via Phase-Aware Placement and DVFS](https://mlsanity.com/paper/2602.18755/) (2026)
- [Watt Counts: Energy-Aware Benchmark for Sustainable LLM Inference on Heterogeneous GPU Architectures](https://mlsanity.com/paper/2604.09048/) (2026)
- [Frontier: Towards Comprehensive and Accurate LLM Inference Simulation](https://mlsanity.com/paper/2605.21312/) (2026)
- [BestServe: Serving Strategies with Optimal Goodput in Collocation and Disaggregation Architectures](https://mlsanity.com/paper/2506.05871/) (2025)
- [FleetOpt: Analytical Fleet Provisioning for LLM Inference with Compress-and-Route as Implementation Mechanism](https://mlsanity.com/paper/2603.16514/) (2026)
- [From Tokens to Watt-hours: Analytical Energy Estimation for LLM Inference on Modern GPUs](https://mlsanity.com/paper/2607.26571/) (2026)
- [VoltanaLLM: Energy-Efficient and SLO-Aware Disaggregated LLM Serving via Adaptive Frequency Control and State-Space Routing](https://mlsanity.com/paper/2509.04827/) (2025)
- [Analytical Provisioning for Attention-FFN Disaggregated LLM Serving under Stochastic Workloads](https://mlsanity.com/paper/2601.21351/) (2026)
- [GPU-to-Grid: Voltage Regulation via GPU Utilization Control](https://mlsanity.com/paper/2602.05116/) (2026)
- [inference-fleet-sim: A Queueing-Theory-Grounded Fleet Capacity Planner for LLM Inference](https://mlsanity.com/paper/2603.16054/) (2026)
- [A Queueing-Theoretic Framework for Stability Analysis of LLM Inference with KV Cache Memory Constraints](https://mlsanity.com/paper/2605.04595/) (2026)
- [PowerSlider: Exploiting Phase Asymmetry for LLM Serving under Demand Response](https://mlsanity.com/paper/2608.21719/) (2026)
- [Mitigation of Datacenter Demand Ramping and Fluctuation using Hybrid ESS and Supercapacitor](https://mlsanity.com/paper/2512.08076/) (2025)
- [SLO-Aware Compute Resource Allocation for Prefill-Decode Disaggregated LLM Inference](https://mlsanity.com/paper/2603.04716/) (2026)
- [From Tokens to Energy Flexibility: Quantization-Enabled Demand Response for Data Centers with LLM Inference Workloads](https://mlsanity.com/paper/2606.18851/) (2026)

## Abstract (arXiv)

Power availability increasingly constrains the operation of AI inference fleets, creating a need for provisioning methods that jointly consider serving capacity and power consumption. Prefill--decode (PD) disaggregation has emerged as a prevalent architecture for large-scale inference serving. However, determining the appropriate numbers of prefill and decode instances is challenging because serving capacity depends jointly on workload characteristics, hardware constraints, queueing, and KV-cache reservations. Existing approaches largely rely on profiling and simulation, providing limited analytical insight into how provisioning decisions shape the tradeoff between serving capacity and power consumption. This paper develops an analytical framework for power-aware provisioning of PD-disaggregated AI inference. Given an inference workload and hardware, the framework models the serving capacity and average power consumption of a provisioned deployment. The serving-capacity model is derived from the joint distribution of input--output lengths and hardware compute and memory limits. In particular, it explicitly captures the coupling between prefill and decode induced by KV-cache reservations, as well as the impact of request queueing. On this basis, the power model determines per-instance power consumption as a function of normalized serving throughput. Together, the models determine the serving capacity--power Pareto front among candidate provisioned deployments, enabling the service provider to choose a provisioned deployment as the workload or available power changes.