Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2601.08833

Revisiting Disaggregated Large Language Model Serving for Performance and Energy Implications

Jiaxi Li·Yue Zhu·Eun Kyung Lee·Klara Nahrstedt
View original source ↗Read source document ↗Export metadata ↓

Abstract

Different from traditional Large Language Model (LLM) serving that colocates the prefill and decode stages on the same GPU, disaggregated serving dedicates distinct GPUs to prefill and decode workload. Once the prefill GPU completes its task, the KV cache must be transferred to the decode GPU. While existing works have proposed various KV cache transfer paths across different memory and storage tiers, there remains a lack of systematic benchmarking that compares their performance and energy efficiency. Meanwhile, although optimization techniques such as KV cache reuse and frequency scaling have been utilized for disaggregated serving, their performance and energy implications have not been rigorously benchmarked. In this paper, we fill this research gap by re-evaluating prefill-decode disaggregation under different KV transfer mediums and optimization strategies. Specifically, we include a new colocated serving baseline and evaluate disaggregated setups under different KV cache transfer paths. Through GPU profiling using dynamic voltage and frequency scaling (DVFS), we identify and compare the performance-energy Pareto frontiers across all setups to evaluate the potential energy savings enabled by disaggregation. Our results show that performance benefits from prefill-decode disaggregation are not guaranteed and depend on the request load and KV transfer mediums. In addition, stage-wise independent frequency scaling enabled by disaggregation does not lead to energy saving due to inherently higher energy consumption of disaggregated serving.

Explore related subjects

cs.PFcs.AIcs.ARcs.DC

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Jiaxi Li, Yue Zhu, Eun Kyung Lee, Klara Nahrstedt. 2025-11-14. Revisiting Disaggregated Large Language Model Serving for Performance and Energy Implications. https://arxiv.org/abs/2601.08833

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Performance Analysis of Low-Order, GPU-accelerated Finite Element Kernels using Kokkos

We study performance portability for low-order, matrix-free finite element kernels, using the example of a vectorial, variable-coefficient PDE operator originating in geophysical models. Written in Kokkos, the kernel is compared on NVIDIA H100, AMD MI250X, AMD MI300A and Intel PVC Max 1550 GPUs. Owing to its low order and to optimizations that reduce the arithmetic, the kernel has a low arithmetic intensity, so that its performance is determined by how the finite element assembly is mapped onto the memory hierarchy. This is a dimension in which the architectures differ even within one vendor family, causing different performance characteristics. We examine how Kokkos' hierarchical parallelism and shared scratch memory, which are used for the shared degrees of freedom of the conforming discretization, behave on each device. Finally, we show how portability gaps can be narrowed with tuning levers such as the size of the thread groups, the balance between occupancy and register use, and the atomic accumulation strategy at the end of the kernel.

cs.PF↗
arXiv2026

Measured Joules, Learned Routes: Learning to Route for Energy-Efficient LLM Serving

Large language models (LLMs) and agentic AI systems are creating rapidly growing inference energy demands as model sizes grow and reasoning trajectories extend. While in practice, many queries do not require the capabilities of the largest available model, and routinely directing such queries to a high-capability model can introduce unnecessary, considerable computation and energy consumption. In this paper, we investigate whether adaptive routing across a heterogeneous pool of LLMs can reduce this energy burden without substantially compromising task performance. We design a language-model-based router that reads in each query and selects an answer model from a fixed candidate pool. The candidate models are first profiled through an offline tournament that records their correctness, latency, power, and GPU energy for each query. Using these measurements, the router is trained through supervised fine-tuning followed by group relative policy optimization (GRPO) with the tailored paradigms. Results demonstrate that learned routing can selectively allocate expensive model capacity based on query context and improve the accuracy-energy tradeoff in multi-LLM serving. Across seven benchmark tasks, we also observe a sharp accuracy-energy phase transition among routers, providing practical insights into improving energy efficiency while maintaining LLM performance.

cs.PF↗
arXiv2026

PrefixBench-H100: Characterizing Prefix Reuse and Time-to-First-Token in H100 LLM Serving

Repeated prompt prefixes are increasingly common in LLM serving workloads, appearing in system prompts, templated retrieval-augmented generation pipelines, agent frameworks, and multi-turn conversations. Modern inference runtimes such as vLLM and TensorRT-LLM provide mechanisms for reusing previously computed KV-cache state across requests, yet it remains unclear when prefix reuse materially improves serving performance on contemporary accelerators and when its benefits are limited by scheduling, cache granularity, concurrency, or memory pressure. This paper presents PrefixBench-H100, a reproducible benchmark and measurement framework for characterizing prefix reuse on a single NVIDIA H100. PrefixBench-H100 combines controlled synthetic traces with chat-style and retrieval-style workloads, and evaluates two widely used LLM serving runtimes under matched workload conditions. The benchmark varies shared-prefix length, suffix diversity, request arrival pattern, concurrency, output length, and cache configuration, while collecting time-to-first-token, inter-token latency, end-to-end latency, throughput, cache-hit statistics, GPU memory usage, and selected profiling traces. The goal of PrefixBench-H100 is not to introduce a new caching algorithm, but to expose the practical operating envelope of prefix reuse for H100-class LLM serving. The study identifies the regime where prefix reuse provides substantial first-token latency reductions and the regime where cache pressure erodes them, while showing that cache effectiveness itself is largely insensitive to concurrency and output length; the cross-runtime differences that remain arise above the cache, in the scheduling layer.

cs.PF↗