Saved in:
Bibliographic Details
Main Authors: Chen, June, Xu, Neal, Huang, Gragas, Zhou, Bok, Liu, Stephen
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.20655
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908793553551360
author Chen, June
Xu, Neal
Huang, Gragas
Zhou, Bok
Liu, Stephen
author_facet Chen, June
Xu, Neal
Huang, Gragas
Zhou, Bok
Liu, Stephen
contents The rapid growth of AI-generated content (AIGC) has enabled high-quality creative production across diverse domains, yet existing systems face critical inefficiencies in throughput, resource utilization, and scalability under concurrent workloads. This paper introduces OnePiece, a large-scale distributed inference system with RDMA optimized for multi-stage AIGC workflows. By decomposing pipelines into fine-grained microservices and leveraging one-sided RDMA communication, OnePiece significantly reduces inter-node latency and CPU overhead while improving GPU utilization. The system incorporates a novel double-ring buffer design to resolve deadlocks in RDMA-aware memory access without CPU involvement. Additionally, a dynamic Node Manager allocates resources elastically across workflow stages in response to real-time load. Experimental results demonstrate that OnePiece reduces GPU resource consumption by 16x in Wan2.1 image-to-video generation compared to monolithic inference pipelines, offering a scalable, fault-tolerant, and efficient solution for production AIGC environments.
format Preprint
id arxiv_https___arxiv_org_abs_2601_20655
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle OnePiece: A Large-Scale Distributed Inference System with RDMA for Complex AI-Generated Content (AIGC) Workflows
Chen, June
Xu, Neal
Huang, Gragas
Zhou, Bok
Liu, Stephen
Distributed, Parallel, and Cluster Computing
The rapid growth of AI-generated content (AIGC) has enabled high-quality creative production across diverse domains, yet existing systems face critical inefficiencies in throughput, resource utilization, and scalability under concurrent workloads. This paper introduces OnePiece, a large-scale distributed inference system with RDMA optimized for multi-stage AIGC workflows. By decomposing pipelines into fine-grained microservices and leveraging one-sided RDMA communication, OnePiece significantly reduces inter-node latency and CPU overhead while improving GPU utilization. The system incorporates a novel double-ring buffer design to resolve deadlocks in RDMA-aware memory access without CPU involvement. Additionally, a dynamic Node Manager allocates resources elastically across workflow stages in response to real-time load. Experimental results demonstrate that OnePiece reduces GPU resource consumption by 16x in Wan2.1 image-to-video generation compared to monolithic inference pipelines, offering a scalable, fault-tolerant, and efficient solution for production AIGC environments.
title OnePiece: A Large-Scale Distributed Inference System with RDMA for Complex AI-Generated Content (AIGC) Workflows
topic Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2601.20655