Online Scheduling for LLM Inference with KV Cache Constraints

• URL: http://arxiv.org/abs/2502.07115v4
• Date: Tue, 20 May 2025 16:29:52 GMT
• Title: Online Scheduling for LLM Inference with KV Cache Constraints
• Authors: Patrick Jaillet, Jiashuo Jiang, Konstantina Mellou, Marco Molinaro, Chara Podimata, Zijie Zhou,
• Abstract summary: Large Language Model (LLM) inference is an intensive process requiring efficient scheduling to optimize latency and resource utilization.<n>We propose a novel and theoretically scheduling algorithm that minimizes inference latency while effectively managing the KV cache's memory.
• Score: 22.133592174540052
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Model (LLM) inference, where a trained model generates text one word at a time in response to user prompts, is a computationally intensive process requiring efficient scheduling to optimize latency and resource utilization. A key challenge in LLM inference is the management of the Key-Value (KV) cache, which reduces redundant computations but introduces memory constraints. In this work, we model LLM inference with KV cache constraints theoretically and propose a novel batching and scheduling algorithm that minimizes inference latency while effectively managing the KV cache's memory. More specifically, we make the following contributions. First, to evaluate the performance of online algorithms for scheduling in LLM inference, we introduce a hindsight optimal benchmark, formulated as an integer program that computes the minimum total inference latency under full future information. Second, we prove that no deterministic online algorithm can achieve a constant competitive ratio when the arrival process is arbitrary. Third, motivated by the computational intractability of solving the integer program at scale, we propose a polynomial-time online scheduling algorithm and show that under certain conditions it can achieve a constant competitive ratio. We also demonstrate our algorithm's strong empirical performance by comparing it to the hindsight optimal in a synthetic dataset. Finally, we conduct empirical evaluations on a real-world public LLM inference dataset, simulating the Llama2-70B model on A100 GPUs, and show that our algorithm significantly outperforms the benchmark algorithms. Overall, our results offer a path toward more sustainable and cost-effective LLM deployment.

Related papers

• Optimizing LLM Inference: Fluid-Guided Online Scheduling with Memory Constraints [14.341123057506827]
  Large Language Models (LLMs) are indispensable in today's applications, but their inference procedure demands significant computational resources. This paper formulates LLM inference optimization as a multi-stage online scheduling problem. We develop a fluid dynamics approximation to provide a tractable benchmark that guides algorithm design.
  arXiv  Detail & Related papers  (2025-04-15T16:00:21Z)
• Seesaw: High-throughput LLM Inference via Model Re-sharding [8.840996987380484]
  We present Seesaw, an inference engine optimized for throughput-oriented tasks. Key idea behind Seesaw is dynamic model re-sharding, a technique that facilitates the dynamic reconfiguration of parallelization strategies.
  arXiv  Detail & Related papers  (2025-03-09T04:14:06Z)
• In-context Demonstration Matters: On Prompt Optimization for Pseudo-Supervision Refinement [71.60563181678323]
  Large language models (LLMs) have achieved great success across diverse tasks, and fine-tuning is sometimes needed to further enhance generation quality.<n>To handle these challenges, a direct solution is to generate high-confidence'' data from unsupervised downstream tasks.<n>We propose a novel approach, pseudo-supervised demonstrations aligned prompt optimization (PAPO) algorithm, which jointly refines both the prompt and the overall pseudo-supervision.
  arXiv  Detail & Related papers  (2024-10-04T03:39:28Z)
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters [27.656263126925815]
  We study the scaling of inference-time computation in LLMs. We find that in both cases, the effectiveness of different approaches to scaling test-time compute critically varies depending on the difficulty of the prompt.
  arXiv  Detail & Related papers  (2024-08-06T17:35:05Z)
• MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention [36.49445805074941]
  MInference (Milliontokens Inference) is a sparse calculation method designed to accelerate pre-filling of long-sequence processing. We demonstrate that MInference effectively reduces inference latency by up to 10x for pre-filling on an A100, while maintaining accuracy.
  arXiv  Detail & Related papers  (2024-07-02T17:59:56Z)
• Bypass Back-propagation: Optimization-based Structural Pruning for Large Language Models via Policy Gradient [57.9629676017527]
  We propose an optimization-based structural pruning on Large-Language Models. We learn the pruning masks in a probabilistic space directly by optimizing the loss of the pruned model. Our method operates for 2.7 hours with around 35GB memory for the 13B models on a single A100 GPU.
  arXiv  Detail & Related papers  (2024-06-15T09:31:03Z)
• Switchable Decision: Dynamic Neural Generation Networks [98.61113699324429]
  We propose a switchable decision to accelerate inference by dynamically assigning resources for each data instance. Our method benefits from less cost during inference while keeping the same accuracy.
  arXiv  Detail & Related papers  (2024-05-07T17:44:54Z)
• Latency-aware Unified Dynamic Networks for Efficient Image Recognition [72.8951331472913]
  LAUDNet is a framework to bridge the theoretical and practical efficiency gap in dynamic networks. It integrates three primary dynamic paradigms-spatially adaptive computation, dynamic layer skipping, and dynamic channel skipping. It can notably reduce the latency of models like ResNet by over 50% on platforms such as V100,3090, and TX2 GPUs.
  arXiv  Detail & Related papers  (2023-08-30T10:57:41Z)
• Accelerating Exact Combinatorial Optimization via RL-based Initialization -- A Case Study in Scheduling [1.3053649021965603]
  This research aims to develop an innovative approach that employs machine learning (ML) for addressing optimization problems. We introduce a novel two-phase RL-to-ILP scheduling framework, which includes three steps: 1) solver as coarse-grain scheduler, 2) solution relaxation and 3) exact solving via ILP. Our framework demonstrates the same scheduling performance compared with using exact scheduling methods while achieving up to 128 $times$ speed improvements.
  arXiv  Detail & Related papers  (2023-08-19T15:52:43Z)
• Maximize to Explore: One Objective Function Fusing Estimation, Planning, and Exploration [87.53543137162488]
  We propose an easy-to-implement online reinforcement learning (online RL) framework called textttMEX. textttMEX integrates estimation and planning components while balancing exploration exploitation automatically. It can outperform baselines by a stable margin in various MuJoCo environments with sparse rewards.
  arXiv  Detail & Related papers  (2023-05-29T17:25:26Z)
• Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline [22.08897444328099]
  Large language models (LLMs) have revolutionized the field of AI, demonstrating unprecedented capacity across various tasks. In this paper, we propose an efficient LLM inference pipeline that harnesses the power of LLMs.
  arXiv  Detail & Related papers  (2023-05-22T15:36:06Z)
• SatLM: Satisfiability-Aided Language Models Using Declarative Prompting [68.40726892904286]
  We propose a new satisfiability-aided language modeling (SatLM) approach for improving the reasoning capabilities of large language models (LLMs) We use an LLM to generate a declarative task specification rather than an imperative program and leverage an off-the-shelf automated theorem prover to derive the final answer. We evaluate SATLM on 8 different datasets and show that it consistently outperforms program-aided LMs in the imperative paradigm.
  arXiv  Detail & Related papers  (2023-05-16T17:55:51Z)
• RESPECT: Reinforcement Learning based Edge Scheduling on Pipelined Coral Edge TPUs [12.952987240366781]
  This work presents a reinforcement learning (RL) based scheduling framework, which learns the behaviors of optimal optimization algorithms. RL generates near-optimal scheduling results with short solving runtime overhead. Our framework has demonstrated up to $sim2.5times$ real-world on-chip runtime inference speedups over the commercial compiler.
  arXiv  Detail & Related papers  (2023-04-10T17:22:12Z)
• Making Linear MDPs Practical via Contrastive Representation Learning [101.75885788118131]
  It is common to address the curse of dimensionality in Markov decision processes (MDPs) by exploiting low-rank representations. We consider an alternative definition of linear MDPs that automatically ensures normalization while allowing efficient representation learning. We demonstrate superior performance over existing state-of-the-art model-based and model-free algorithms on several benchmarks.
  arXiv  Detail & Related papers  (2022-07-14T18:18:02Z)
• Accelerating Deep Learning Classification with Error-controlled Approximate-key Caching [72.50506500576746]
  We propose a novel caching paradigm, that we named approximate-key caching. While approximate cache hits alleviate DL inference workload and increase the system throughput, they however introduce an approximation error. We analytically model our caching system performance for classic LRU and ideal caches, we perform a trace-driven evaluation of the expected performance, and we compare the benefits of our proposed approach with the state-of-the-art similarity caching.
  arXiv  Detail & Related papers  (2021-12-13T13:49:11Z)
• AsySQN: Faster Vertical Federated Learning Algorithms with Better Computation Resource Utilization [159.75564904944707]
  We propose an asynchronous quasi-Newton (AsySQN) framework for vertical federated learning (VFL) The proposed algorithms make descent steps scaled by approximate without calculating the inverse Hessian matrix explicitly. We show that the adopted asynchronous computation can make better use of the computation resource.
  arXiv  Detail & Related papers  (2021-09-26T07:56:10Z)
• Adaptive Sampling for Best Policy Identification in Markov Decision Processes [79.4957965474334]
  We investigate the problem of best-policy identification in discounted Markov Decision (MDPs) when the learner has access to a generative model. The advantages of state-of-the-art algorithms are discussed and illustrated.
  arXiv  Detail & Related papers  (2020-09-28T15:22:24Z)
• Combining Deep Learning and Optimization for Security-Constrained Optimal Power Flow [94.24763814458686]
  Security-constrained optimal power flow (SCOPF) is fundamental in power systems. Modeling of APR within the SCOPF problem results in complex large-scale mixed-integer programs. This paper proposes a novel approach that combines deep learning and robust optimization techniques.
  arXiv  Detail & Related papers  (2020-07-14T12:38:21Z)

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.