Improving the End-to-End Efficiency of Offline Inference for Multi-LLM Applications Based on Sampling and Simulation

• URL: http://arxiv.org/abs/2503.16893v1
• Date: Fri, 21 Mar 2025 06:56:35 GMT
• Title: Improving the End-to-End Efficiency of Offline Inference for Multi-LLM Applications Based on Sampling and Simulation
• Authors: Jingzhi Fang, Yanyan Shen, Yue Wang, Lei Chen,
• Abstract summary: We aim to improve the offline end-to-end inference efficiency of multi-LLM applications in a single-node multi-GPU environment.<n>We propose a sampling-then-simulation method to estimate the model running time.<n>Experiments on 3 applications and a mixed application show that SamuLLM can achieve 1.0-2.4$times$ end-to-end speedups.
• Score: 23.318601470116498
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: As large language models (LLMs) have shown great success in many tasks, they are used in various applications. While a lot of works have focused on the efficiency of single-LLM application (e.g., offloading, request scheduling, parallelism strategy selection), multi-LLM applications receive less attention, particularly in offline inference scenarios. In this work, we aim to improve the offline end-to-end inference efficiency of multi-LLM applications in the single-node multi-GPU environment. The problem involves two key decisions: (1) determining which LLMs to run concurrently each time (we may not run all the models at the same time), and (2) selecting a parallelism strategy to use for each LLM. This problem is NP-hard. Naive solutions may not work well because the running time for a model to complete a set of requests depends on the request workload and the selected parallelism strategy, and they lack an accurate model of the running time. As the LLM output lengths are unknown before running, to estimate the model running time, we propose a sampling-then-simulation method which first estimates the output lengths by sampling from an empirical cumulative function we obtained from a large dataset in advance, and then simulates the LLM inference process accordingly. Based on the simulation, we estimate the per-iteration latencys to get the total latency. A greedy method is proposed to optimize the scheduling of the LLMs in the application across the GPUs. We then propose a framework SamuLLM which contains two phases: planning, which calls the greedy method for an application and running, which runs the application and dynamically adjust the model scheduling based on the runtime information. Experiments on 3 applications and a mixed application show that SamuLLM can achieve 1.0-2.4$\times$ end-to-end speedups compared to the competitors.

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)
• Optimizing Model Selection for Compound AI Systems [76.69936664916061]
  We propose an efficient framework for model selection in compound systems.<n>It iteratively selects one module and allocates to it the model with the highest module-wise performance.<n>It confers 5%-70% accuracy gains compared to using the same LLM for all modules.
  arXiv  Detail & Related papers  (2025-02-20T18:36:25Z)
• Autellix: An Efficient Serving Engine for LLM Agents as General Programs [59.673243129044465]
  Large language model (LLM) applications are evolving beyond simple chatbots into dynamic, general-purpose agentic programs.<n>Existing LLM serving systems ignore dependencies between programs and calls, missing significant opportunities for optimization.<n>We introduce Autellix, an LLM serving system that treats programs as first-class citizens to minimize their end-to-end latencies.
  arXiv  Detail & Related papers  (2025-02-19T18:59:30Z)
• Multi-Bin Batching for Increasing LLM Inference Throughput [19.652542432683234]
  Large language models (LL) grow in popularity improving the efficiency of their systems.<n> requests is a critical step in scheduling jobs on servers.<n> requests often have varying generation lengths, causing resource underutilization.<n>We formalize this problem from a queueing-theoretic perspective, and aim to design a throughput control policy.
  arXiv  Detail & Related papers  (2024-12-03T03:16:12Z)
• Is the GPU Half-Empty or Half-Full? Practical Scheduling Techniques for LLMs [3.7758841366694353]
  We survey scheduling techniques from the literature and from practical serving systems. We find that schedulers from the literature often achieve good performance but introduce significant complexity. In contrast, schedulers in practical deployments often leave easy performance gains on the table but are easy to implement, deploy and configure.
  arXiv  Detail & Related papers  (2024-10-23T13:05:46Z)
• EPS-MoE: Expert Pipeline Scheduler for Cost-Efficient MoE Inference [49.94169109038806]
  This paper introduces EPS-MoE, a novel expert pipeline scheduler for MoE that surpasses the existing parallelism schemes.<n>Our results demonstrate at most 52.4% improvement in prefill throughput compared to existing parallel inference methods.
  arXiv  Detail & Related papers  (2024-10-16T05:17:49Z)
• QPO: Query-dependent Prompt Optimization via Multi-Loop Offline Reinforcement Learning [58.767866109043055]
  We introduce Query-dependent Prompt Optimization (QPO), which iteratively fine-tune a small pretrained language model to generate optimal prompts tailored to the input queries. We derive insights from offline prompting demonstration data, which already exists in large quantities as a by-product of benchmarking diverse prompts on open-sourced tasks. Experiments on various LLM scales and diverse NLP and math tasks demonstrate the efficacy and cost-efficiency of our method in both zero-shot and few-shot scenarios.
  arXiv  Detail & Related papers  (2024-08-20T03:06:48Z)
• Teola: Towards End-to-End Optimization of LLM-based Applications [13.478509565946354]
  Large language model (LLM)-based applications contribute to the end-to-end latency.<n>Existing frameworks employ coarse-grained orchestration with task modules, which confines optimizations to within each module.<n>We propose fine-grained end-to-end orchestration, which utilizes task primitives as the basic units and represents each query's workflow as a primitive-level dataflow graph.
  arXiv  Detail & Related papers  (2024-06-29T05:59:53Z)
• Preble: Efficient Distributed Prompt Scheduling for LLM Serving [8.706905652975554]
  This paper proposes Preble, the first distributed LLM serving platform that targets and optimize for prompt sharing. We designed a distributed scheduling system that co-optimizes KV state reuse and computation load-balancing with a new scheduling algorithm and a hierarchical scheduling mechanism. Our evaluation of Preble with real workloads and request arrival patterns on two open-source LLMs shows that Preble outperforms the SOTA serving systems by 1.5X to 14.5X on average latency and 2X to 10X on p99 latency.
  arXiv  Detail & Related papers  (2024-05-08T06:30:58Z)
• Efficient Interactive LLM Serving with Proxy Model-based Sequence Length Prediction [8.705908108054878]
  Large models (LLMs) have been driving a new wave of AI applications across numerous domains. We present a speculative shortest-job-first (SSJF) scheduler that uses a light proxy model to predict LLM output sequence lengths.
  arXiv  Detail & Related papers  (2024-04-12T14:46:15Z)
• Are More LLM Calls All You Need? Towards Scaling Laws of Compound Inference Systems [76.69936664916061]
  We study how the number of LM calls affects the performance of Vote and Filter-Vote. We find, surprisingly, that across multiple language tasks, the performance of both Vote and Filter-Vote can first increase but then decrease as a function of the number of LM calls.
  arXiv  Detail & Related papers  (2024-03-04T19:12:48Z)
• Graph-enhanced Large Language Models in Asynchronous Plan Reasoning [18.402877904882107]
  We find that large language models (LLMs) behave poorly when not supplied with illustrations about the task-solving process in our benchmark AsyncHow. We propose a novel technique called Plan Like a Graph (PLaG) that combines graphs with natural language prompts and achieves state-of-the-art results.
  arXiv  Detail & Related papers  (2024-02-05T08:26:33Z)
• RMP-SAM: Towards Real-Time Multi-Purpose Segment Anything [117.02741621686677]
  This work explores a novel real-time segmentation setting called real-time multi-purpose segmentation. It contains three fundamental sub-tasks: interactive segmentation, panoptic segmentation, and video instance segmentation. We present a novel dynamic convolution-based method, Real-Time Multi-Purpose SAM (RMP-SAM) It contains an efficient encoder and an efficient decoupled adapter to perform prompt-driven decoding.
  arXiv  Detail & Related papers  (2024-01-18T18:59:30Z)

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.