Efficient Unified Caching for Accelerating Heterogeneous AI Workloads

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wang, Tianze, Liu, Yifei, Chen, Chen, Zuo, Pengfei, Zhang, Jiawei, Weng, Qizhen, Chen, Yin, Han, Zhenhua, Zhao, Jieru, Chen, Quan, Guo, Minyi
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909649110827008
author Wang, Tianze
Liu, Yifei
Chen, Chen
Zuo, Pengfei
Zhang, Jiawei
Weng, Qizhen
Chen, Yin
Han, Zhenhua
Zhao, Jieru
Chen, Quan
Guo, Minyi
author_facet Wang, Tianze
Liu, Yifei
Chen, Chen
Zuo, Pengfei
Zhang, Jiawei
Weng, Qizhen
Chen, Yin
Han, Zhenhua
Zhao, Jieru
Chen, Quan
Guo, Minyi
contents Modern AI clusters, which host diverse workloads like data pre-processing, training and inference, often store the large-volume data in cloud storage and employ caching frameworks to facilitate remote data access. To avoid code-intrusion complexity and minimize cache space wastage, it is desirable to maintain a unified cache shared by all the workloads. However, existing cache management strategies, designed for specific workloads, struggle to handle the heterogeneous AI workloads in a cluster -- which usually exhibit heterogeneous access patterns and item storage granularities. In this paper, we propose IGTCache, a unified, high-efficacy cache for modern AI clusters. IGTCache leverages a hierarchical access abstraction, AccessStreamTree, to organize the recent data accesses in a tree structure, facilitating access pattern detection at various granularities. Using this abstraction, IGTCache applies hypothesis testing to categorize data access patterns as sequential, random, or skewed. Based on these detected access patterns and granularities, IGTCache tailors optimal cache management strategies including prefetching, eviction, and space allocation accordingly. Experimental results show that IGTCache increases the cache hit ratio by 55.6% over state-of-the-art caching frameworks, reducing the overall job completion time by 52.2%.
format Preprint
id arxiv_https___arxiv_org_abs_2506_12370
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Efficient Unified Caching for Accelerating Heterogeneous AI Workloads
Wang, Tianze
Liu, Yifei
Chen, Chen
Zuo, Pengfei
Zhang, Jiawei
Weng, Qizhen
Chen, Yin
Han, Zhenhua
Zhao, Jieru
Chen, Quan
Guo, Minyi
Distributed, Parallel, and Cluster Computing
Machine Learning
Modern AI clusters, which host diverse workloads like data pre-processing, training and inference, often store the large-volume data in cloud storage and employ caching frameworks to facilitate remote data access. To avoid code-intrusion complexity and minimize cache space wastage, it is desirable to maintain a unified cache shared by all the workloads. However, existing cache management strategies, designed for specific workloads, struggle to handle the heterogeneous AI workloads in a cluster -- which usually exhibit heterogeneous access patterns and item storage granularities. In this paper, we propose IGTCache, a unified, high-efficacy cache for modern AI clusters. IGTCache leverages a hierarchical access abstraction, AccessStreamTree, to organize the recent data accesses in a tree structure, facilitating access pattern detection at various granularities. Using this abstraction, IGTCache applies hypothesis testing to categorize data access patterns as sequential, random, or skewed. Based on these detected access patterns and granularities, IGTCache tailors optimal cache management strategies including prefetching, eviction, and space allocation accordingly. Experimental results show that IGTCache increases the cache hit ratio by 55.6% over state-of-the-art caching frameworks, reducing the overall job completion time by 52.2%.
title Efficient Unified Caching for Accelerating Heterogeneous AI Workloads
topic Distributed, Parallel, and Cluster Computing
Machine Learning
url https://arxiv.org/abs/2506.12370