PPipe: Efficient Video Analytics Serving on Heterogeneous GPU Clusters via Pool-Based Pipeline Parallelism

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kong, Z. Jonny, Xu, Qiang, Hu, Y. Charlie
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918103852515328
author Kong, Z. Jonny
Xu, Qiang
Hu, Y. Charlie
author_facet Kong, Z. Jonny
Xu, Qiang
Hu, Y. Charlie
contents With the rapid innovation of GPUs, heterogeneous GPU clusters in both public clouds and on-premise data centers have become increasingly commonplace. In this paper, we demonstrate how pipeline parallelism, a technique wellstudied for throughput-oriented deep learning model training, can be used effectively for serving latency-bound model inference, e.g., in video analytics systems, on heterogeneous GPU clusters. Our work exploits the synergy between diversity in model layers and diversity in GPU architectures, which results in comparable inference latency for many layers when running on low-class and high-class GPUs. We explore how such overlooked capability of low-class GPUs can be exploited using pipeline parallelism and present a novel inference serving system, PPipe, that employs pool-based pipeline parallelism via an MILP-based control plane and a data plane that performs resource reservation-based adaptive batching. Evaluation results on diverse workloads (18 CNN models) show that PPipe achieves 41.1% - 65.5% higher utilization of low-class GPUs while maintaining high utilization of high-class GPUs, leading to 32.2% - 75.1% higher serving throughput compared to various baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2507_18748
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PPipe: Efficient Video Analytics Serving on Heterogeneous GPU Clusters via Pool-Based Pipeline Parallelism
Kong, Z. Jonny
Xu, Qiang
Hu, Y. Charlie
Distributed, Parallel, and Cluster Computing
With the rapid innovation of GPUs, heterogeneous GPU clusters in both public clouds and on-premise data centers have become increasingly commonplace. In this paper, we demonstrate how pipeline parallelism, a technique wellstudied for throughput-oriented deep learning model training, can be used effectively for serving latency-bound model inference, e.g., in video analytics systems, on heterogeneous GPU clusters. Our work exploits the synergy between diversity in model layers and diversity in GPU architectures, which results in comparable inference latency for many layers when running on low-class and high-class GPUs. We explore how such overlooked capability of low-class GPUs can be exploited using pipeline parallelism and present a novel inference serving system, PPipe, that employs pool-based pipeline parallelism via an MILP-based control plane and a data plane that performs resource reservation-based adaptive batching. Evaluation results on diverse workloads (18 CNN models) show that PPipe achieves 41.1% - 65.5% higher utilization of low-class GPUs while maintaining high utilization of high-class GPUs, leading to 32.2% - 75.1% higher serving throughput compared to various baselines.
title PPipe: Efficient Video Analytics Serving on Heterogeneous GPU Clusters via Pool-Based Pipeline Parallelism
topic Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2507.18748