semi-PD: Towards Efficient LLM Serving via Phase-Wise Disaggregated Computation and Unified Storage

• URL: http://arxiv.org/abs/2504.19867v1
• Date: Mon, 28 Apr 2025 15:00:03 GMT
• Title: semi-PD: Towards Efficient LLM Serving via Phase-Wise Disaggregated Computation and Unified Storage
• Authors: Ke Hong, Lufang Chen, Zhong Wang, Xiuhong Li, Qiuli Mao, Jianping Ma, Chao Xiong, Guanyu Wu, Buhe Han, Guohao Dai, Yun Liang, Yu Wang,
• Abstract summary: We propose a novel large language model (LLM) serving system, semi-PD, characterized by disaggregated computation and unified storage.<n>Compared to state-of-the-art systems, semi-PD maintains lower latency at higher request rates, reducing the average end-to-end latency per request by 1.27-2.58x.
• Score: 6.805644270436825
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Existing large language model (LLM) serving systems fall into two categories: 1) a unified system where prefill phase and decode phase are co-located on the same GPU, sharing the unified computational resource and storage, and 2) a disaggregated system where the two phases are disaggregated to different GPUs. The design of the disaggregated system addresses the latency interference and sophisticated scheduling issues in the unified system but leads to storage challenges including 1) replicated weights for both phases that prevent flexible deployment, 2) KV cache transfer overhead between the two phases, 3) storage imbalance that causes substantial wasted space of the GPU capacity, and 4) suboptimal resource adjustment arising from the difficulties in migrating KV cache. Such storage inefficiency delivers poor serving performance under high request rates. In this paper, we identify that the advantage of the disaggregated system lies in the disaggregated computation, i.e., partitioning the computational resource to enable the asynchronous computation of two phases. Thus, we propose a novel LLM serving system, semi-PD, characterized by disaggregated computation and unified storage. In semi-PD, we introduce a computation resource controller to achieve disaggregated computation at the streaming multi-processor (SM) level, and a unified memory manager to manage the asynchronous memory access from both phases. semi-PD has a low-overhead resource adjustment mechanism between the two phases, and a service-level objective (SLO) aware dynamic partitioning algorithm to optimize the SLO attainment. Compared to state-of-the-art systems, semi-PD maintains lower latency at higher request rates, reducing the average end-to-end latency per request by 1.27-2.58x on DeepSeek series models, and serves 1.55-1.72x more requests adhering to latency constraints on Llama series models.

Related papers

• Apt-Serve: Adaptive Request Scheduling on Hybrid Cache for Scalable LLM Inference Serving [22.66354939370058]
  Apt-Serve is a framework designed to enhance effective throughput in large language model (LLM) inference serving systems.<n>A new hybrid cache scheme combines KV cache with a memory-efficient hidden cache for reusable input hidden state vectors, allowing large batch sizes and improving request.<n>We show that Apt-Serve achieves up to 8.8x improvement in effective throughput compared to the state-of-the-art inference serving systems.
  arXiv  Detail & Related papers  (2025-04-10T06:51:23Z)
• Injecting Adrenaline into LLM Serving: Boosting Resource Utilization and Throughput via Attention Disaggregation [23.130886760027586]
  In large language model (LLM) serving systems, executing each request consists of two phases: the compute-intensive prefill phase and the memory-intensive decoding phase.<n>This paper proposes Adrenaline, an attention disaggregation and offloading mechanism designed to enhance resource utilization and performance.<n> Experimental results show that Adrenaline achieves 2.28x higher memory capacity and 2.07x better memory bandwidth utilization in prefill instances.
  arXiv  Detail & Related papers  (2025-03-26T13:48:35Z)
• Joint Transmit and Pinching Beamforming for Pinching Antenna Systems (PASS): Optimization-Based or Learning-Based? [89.05848771674773]
  A novel antenna system ()-enabled downlink multi-user multiple-input single-output (MISO) framework is proposed.<n>It consists of multiple waveguides, which equip numerous low-cost antennas, named (PAs)<n>The positions of PAs can be reconfigured to both spanning large-scale path and space.
  arXiv  Detail & Related papers  (2025-02-12T18:54:10Z)
• MAS-Attention: Memory-Aware Stream Processing for Attention Acceleration on Resource-Constrained Edge Devices [24.1144641404561]
  We propose a scheme for exact attention inference acceleration on memory-constrained edge accelerators.<n>We show up to 2.75x speedup and 54% reduction in energy consumption as compared to the state-of-the-art attention fusion method (FLAT) in the edge computing scenario.
  arXiv  Detail & Related papers  (2024-11-20T19:44:26Z)
• Digital Twin-Assisted Federated Learning with Blockchain in Multi-tier Computing Systems [67.14406100332671]
  In Industry 4.0 systems, resource-constrained edge devices engage in frequent data interactions. This paper proposes a digital twin (DT) and federated digital twin (FL) scheme. The efficacy of our proposed cooperative interference-based FL process has been verified through numerical analysis.
  arXiv  Detail & Related papers  (2024-11-04T17:48:02Z)
• POD-Attention: Unlocking Full Prefill-Decode Overlap for Faster LLM Inference [9.164093249308419]
  We present POD-Attention - the first GPU kernel that efficiently computes attention for hybrid batches.<n> POD-Attention aims to maximize the utilization of both compute and memory bandwidth by carefully allocating the GPU's resources.
  arXiv  Detail & Related papers  (2024-10-23T17:06:56Z)
• Progressive Mixed-Precision Decoding for Efficient LLM Inference [49.05448842542558]
  We introduce Progressive Mixed-Precision Decoding (PMPD) to address the memory-boundedness of decoding.<n>PMPD achieves 1.4$-$12.2$times$ speedup in matrix-vector multiplications over fp16 models.<n>Our approach delivers a throughput gain of 3.8$-$8.0$times$ over fp16 models and up to 1.54$times$ over uniform quantization approaches.
  arXiv  Detail & Related papers  (2024-10-17T11:46:33Z)
• FusionLLM: A Decentralized LLM Training System on Geo-distributed GPUs with Adaptive Compression [55.992528247880685]
  Decentralized training faces significant challenges regarding system design and efficiency. We present FusionLLM, a decentralized training system designed and implemented for training large deep neural networks (DNNs) We show that our system and method can achieve 1.45 - 9.39x speedup compared to baseline methods while ensuring convergence.
  arXiv  Detail & Related papers  (2024-10-16T16:13:19Z)
• Compute Or Load KV Cache? Why Not Both? [6.982874528357836]
  Cake is a novel KV cache loading system that optimally utilizes both computational and I/O resources in parallel. Cake achieves on average 2.6x reduction in Time to First Token (TTFT) compared to compute-only and I/O-only methods.
  arXiv  Detail & Related papers  (2024-10-04T01:11:09Z)
• Joint Service Caching, Communication and Computing Resource Allocation in Collaborative MEC Systems: A DRL-based Two-timescale Approach [15.16859210403316]
  Meeting the strict Quality of Service (QoS) requirements of terminals has imposed a challenge on Multiaccess Edge Computing (MEC) systems. We propose a collaborative framework that facilitates resource sharing between the edge servers. We show that our proposed algorithm outperforms the baseline algorithms in terms of the average switching and cache cost.
  arXiv  Detail & Related papers  (2023-07-19T00:27:49Z)
• Federated Learning for Energy-limited Wireless Networks: A Partial Model Aggregation Approach [79.59560136273917]
  limited communication resources, bandwidth and energy, and data heterogeneity across devices are main bottlenecks for federated learning (FL) We first devise a novel FL framework with partial model aggregation (PMA) The proposed PMA-FL improves 2.72% and 11.6% accuracy on two typical heterogeneous datasets.
  arXiv  Detail & Related papers  (2022-04-20T19:09:52Z)
• Adaptive Subcarrier, Parameter, and Power Allocation for Partitioned Edge Learning Over Broadband Channels [69.18343801164741]
  partitioned edge learning (PARTEL) implements parameter-server training, a well known distributed learning method, in wireless network. We consider the case of deep neural network (DNN) models which can be trained using PARTEL by introducing some auxiliary variables.
  arXiv  Detail & Related papers  (2020-10-08T15:27:50Z)

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.