Efficient Orchestrated AI Workflows Execution on Scale-out Spatial Architecture

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Deng, Jinyi, Tang, Xinru, Yue, Zhiheng, Lu, Guangyang, Yang, Qize, Zhang, Jiahao, Li, Jinxi, Li, Chao, Wei, Shaojun, Hu, Yang, Yin, Shouyi
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916262263652352
author Deng, Jinyi
Tang, Xinru
Yue, Zhiheng
Lu, Guangyang
Yang, Qize
Zhang, Jiahao
Li, Jinxi
Li, Chao
Wei, Shaojun
Hu, Yang
Yin, Shouyi
author_facet Deng, Jinyi
Tang, Xinru
Yue, Zhiheng
Lu, Guangyang
Yang, Qize
Zhang, Jiahao
Li, Jinxi
Li, Chao
Wei, Shaojun
Hu, Yang
Yin, Shouyi
contents Given the increasing complexity of AI applications, traditional spatial architectures frequently fall short. Our analysis identifies a pattern of interconnected, multi-faceted tasks encompassing both AI and general computational processes. In response, we have conceptualized "Orchestrated AI Workflows," an approach that integrates various tasks with logic-driven decisions into dynamic, sophisticated workflows. Specifically, we find that the intrinsic Dual Dynamicity of Orchestrated AI Workflows, namely dynamic execution times and frequencies of Task Blocks, can be effectively represented using the Orchestrated Workflow Graph. Furthermore, the intrinsic Dual Dynamicity poses challenges to existing spatial architecture, namely Indiscriminate Resource Allocation, Reactive Load Rebalancing, and Contagious PEA Idleness. To overcome these challenges, we present Octopus, a scale-out spatial architecture and a suite of advanced scheduling strategies optimized for executing Orchestrated AI Workflows, such as the Discriminate Dual-Scheduling Mechanism, Adaptive TBU Scheduling Strategy, and Proactive Cluster Scheduling Strategy. Our evaluations demonstrate that Octopus significantly outperforms traditional architectures in handling the dynamic demands of Orchestrated AI Workflows, and possesses robust scalability in large scale hardware such as wafer-scale chip.
format Preprint
id arxiv_https___arxiv_org_abs_2405_17221
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Efficient Orchestrated AI Workflows Execution on Scale-out Spatial Architecture
Deng, Jinyi
Tang, Xinru
Yue, Zhiheng
Lu, Guangyang
Yang, Qize
Zhang, Jiahao
Li, Jinxi
Li, Chao
Wei, Shaojun
Hu, Yang
Yin, Shouyi
Artificial Intelligence
Hardware Architecture
Given the increasing complexity of AI applications, traditional spatial architectures frequently fall short. Our analysis identifies a pattern of interconnected, multi-faceted tasks encompassing both AI and general computational processes. In response, we have conceptualized "Orchestrated AI Workflows," an approach that integrates various tasks with logic-driven decisions into dynamic, sophisticated workflows. Specifically, we find that the intrinsic Dual Dynamicity of Orchestrated AI Workflows, namely dynamic execution times and frequencies of Task Blocks, can be effectively represented using the Orchestrated Workflow Graph. Furthermore, the intrinsic Dual Dynamicity poses challenges to existing spatial architecture, namely Indiscriminate Resource Allocation, Reactive Load Rebalancing, and Contagious PEA Idleness. To overcome these challenges, we present Octopus, a scale-out spatial architecture and a suite of advanced scheduling strategies optimized for executing Orchestrated AI Workflows, such as the Discriminate Dual-Scheduling Mechanism, Adaptive TBU Scheduling Strategy, and Proactive Cluster Scheduling Strategy. Our evaluations demonstrate that Octopus significantly outperforms traditional architectures in handling the dynamic demands of Orchestrated AI Workflows, and possesses robust scalability in large scale hardware such as wafer-scale chip.
title Efficient Orchestrated AI Workflows Execution on Scale-out Spatial Architecture
topic Artificial Intelligence
Hardware Architecture
url https://arxiv.org/abs/2405.17221