Efficient Direct-Connect Topologies for Collective Communications

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zhao, Liangyu, Pal, Siddharth, Chugh, Tapan, Wang, Weiyang, Fantl, Jason, Basu, Prithwish, Khoury, Joud, Krishnamurthy, Arvind
Formato: Preprint
Publicado: 2022
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915131707883520
author Zhao, Liangyu
Pal, Siddharth
Chugh, Tapan
Wang, Weiyang
Fantl, Jason
Basu, Prithwish
Khoury, Joud
Krishnamurthy, Arvind
author_facet Zhao, Liangyu
Pal, Siddharth
Chugh, Tapan
Wang, Weiyang
Fantl, Jason
Basu, Prithwish
Khoury, Joud
Krishnamurthy, Arvind
contents We consider the problem of distilling efficient network topologies for collective communications. We provide an algorithmic framework for constructing direct-connect topologies optimized for the latency vs. bandwidth trade-off associated with the workload. Our approach synthesizes many different topologies and schedules for a given cluster size and degree and then identifies the appropriate topology and schedule for a given workload. Our algorithms start from small, optimal base topologies and associated communication schedules and use techniques that can be iteratively applied to derive much larger topologies and schedules. Additionally, we incorporate well-studied large-scale graph topologies into our algorithmic framework by producing efficient collective schedules for them using a novel polynomial-time algorithm. Our evaluation uses multiple testbeds and large-scale simulations to demonstrate significant performance benefits from our derived topologies and schedules.
format Preprint
id arxiv_https___arxiv_org_abs_2202_03356
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Efficient Direct-Connect Topologies for Collective Communications
Zhao, Liangyu
Pal, Siddharth
Chugh, Tapan
Wang, Weiyang
Fantl, Jason
Basu, Prithwish
Khoury, Joud
Krishnamurthy, Arvind
Networking and Internet Architecture
Distributed, Parallel, and Cluster Computing
Machine Learning
We consider the problem of distilling efficient network topologies for collective communications. We provide an algorithmic framework for constructing direct-connect topologies optimized for the latency vs. bandwidth trade-off associated with the workload. Our approach synthesizes many different topologies and schedules for a given cluster size and degree and then identifies the appropriate topology and schedule for a given workload. Our algorithms start from small, optimal base topologies and associated communication schedules and use techniques that can be iteratively applied to derive much larger topologies and schedules. Additionally, we incorporate well-studied large-scale graph topologies into our algorithmic framework by producing efficient collective schedules for them using a novel polynomial-time algorithm. Our evaluation uses multiple testbeds and large-scale simulations to demonstrate significant performance benefits from our derived topologies and schedules.
title Efficient Direct-Connect Topologies for Collective Communications
topic Networking and Internet Architecture
Distributed, Parallel, and Cluster Computing
Machine Learning
url https://arxiv.org/abs/2202.03356