Towards Multi-Model LLM Schedulers: Empirical Insights into Offloading and Preemption

• URL: http://arxiv.org/abs/2605.19593v1
• Date: Tue, 19 May 2026 09:39:16 GMT
• Title: Towards Multi-Model LLM Schedulers: Empirical Insights into Offloading and Preemption
• Abstract summary: We show how different Large Language Models behave across hardware platforms, focusing on the performance implications of layer offloading and preemption.<n>We identify a set of key features that future schedulers must consider, including model-specific offloading sensitivity, workload characteristics, and the cost structure of preemption and data transfer.
• Score: 3.6129975381123107
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Modern deployments of Large Language Models (LLMs) increasingly require serving multiple models with diverse architectures, sizes, and specialization on shared, heterogeneous hardware. This setting introduces new challenges for resource allocation, dispatching, and scheduling, particularly under GPU memory constraints where partial CPU-GPU offloading and preemption become necessary. While existing systems primarily optimize throughput for a single model, comparatively little work addresses multi-model scheduling under these conditions. In this paper, we present an empirical study of how different LLMs behave across hardware platforms, focusing on the performance implications of layer offloading and preemption. We show that offloading leads to strongly non-linear and model-dependent degradation in decode throughput, with smaller models exhibiting sharper sensitivity to reduced GPU residency. We further demonstrate that preemption incurs substantial overhead, largely dominated by model state reload rather than key-value cache transfer, and that this cost varies significantly across models and hardware platforms. Additionally, we highlight the role of sequence length and interconnect bandwidth in amplifying data movement and execution inefficiencies. Based on these findings, we identify a set of key features that future schedulers must consider, including model-specific offloading sensitivity, workload characteristics, and the cost structure of preemption and data transfer. These insights provide guidance for the design of next-generation LLM serving systems capable of efficiently managing heterogeneous, multi-model workloads with hybrid CPU-GPU execution.

Related papers

• M*: A Modular, Extensible, Serving System for Multimodal Models [62.77975969000349]
  We present M*, a universal serving system for efficient serving of composite AI models.<n>We show how it can concisely capture composite models from a broad range of families.<n>M* also outperforms the V-JEPA 2-AC rollout baseline for robotic planning by up to 12.5x.
  arXiv  Detail & Related papers  (2026-06-10T21:22:22Z)
• Stacked from One: Multi-Scale Self-Injection for Context Window Extension [69.24689919827817]
  modelname is a novel framework based on multi-grained context compression and query-aware information acquisition.<n>modelnameachieves performance superior or comparable to strong baselines.
  arXiv  Detail & Related papers  (2026-03-05T03:16:16Z)
• Scaling LLM Speculative Decoding: Non-Autoregressive Forecasting in Large-Batch Scenarios [76.85739138203014]
  We present SpecFormer, a novel architecture that accelerates unidirectional and attention mechanisms.<n>We demonstrate that SpecFormer achieves lower training demands and reduced computational costs.
  arXiv  Detail & Related papers  (2025-11-25T14:20:08Z)
• SVTime: Small Time Series Forecasting Models Informed by "Physics" of Large Vision Model Forecasters [86.38433605933515]
  Time series AI is crucial for analyzing dynamic web content.<n>Given their energy-intensive training, inference, and hardware demands, using large models as a one-fits-all solution raises serious concerns about carbon footprint and sustainability.<n>This paper introduces SVTime, a novel Small model inspired by large Vision model (LVM) forecasters for long-term Time series forecasting (LTSF)
  arXiv  Detail & Related papers  (2025-10-10T18:42:23Z)
• Characterizing the Efficiency of Distributed Training: A Power, Performance, and Thermal Perspective [6.51239603014107]
  Large Language Models (LLMs) have pushed training workloads beyond the limits of single-node analysis.<n>We present a comprehensive characterization of LLM training across diverse real-world workloads and hardware platforms.
  arXiv  Detail & Related papers  (2025-09-12T16:05:07Z)
• Understanding and Optimizing Multi-Stage AI Inference Pipelines [11.254219071373319]
  HERMES is a Heterogeneous Multi-stage LLM inference Execution Simulator.<n> HERMES supports heterogeneous clients executing multiple models concurrently unlike prior frameworks.<n>We explore the impact of reasoning stages on end-to-end latency, optimal strategies for hybrid pipelines, and the architectural implications of remote KV cache retrieval.
  arXiv  Detail & Related papers  (2025-04-14T00:29:49Z)
• PLM: Efficient Peripheral Language Models Hardware-Co-Designed for Ubiquitous Computing [48.30406812516552]
  We introduce the PLM, a Peripheral Language Model, developed through a co-design process that jointly optimize model architecture and edge system constraints.<n>PLM employs a Multi-head Latent Attention mechanism and employs the squared ReLU activation function to encourage sparsity, thereby reducing peak memory footprint.<n> evaluation results demonstrate that PLM outperforms existing small language models trained on publicly available data.
  arXiv  Detail & Related papers  (2025-03-15T15:11:17Z)
• AutoHete: An Automatic and Efficient Heterogeneous Training System for LLMs [68.99086112477565]
  Transformer-based large language models (LLMs) have demonstrated exceptional capabilities in sequence modeling and text generation.<n>Existing heterogeneous training methods significantly expand the scale of trainable models but introduce substantial communication overheads and CPU workloads.<n>We propose AutoHete, an automatic and efficient heterogeneous training system compatible with both single- GPU and multi- GPU environments.
  arXiv  Detail & Related papers  (2025-02-27T14:46:22Z)
• LV-UNet: A Lightweight and Vanilla Model for Medical Image Segmentation [17.839755917342362]
  This paper introduces LV-UNet, a lightweight and vanilla model that leverages pre-trained MobileNetv3-Large backbones and incorporates modules.<n> Experimental results on ISIC 2016, BUSI, CVCClinicDB, CVCColonDB, and KvairSEG datasets demonstrate a better tradeoff between performance and the computational load.
  arXiv  Detail & Related papers  (2024-08-29T20:19:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.