Network-accelerated cluster scheduler

Efficient use of computing clusters is crucial in large-scale data centers: even small gains in utilization can save millions of dollars. However, as the number of microsecond-scale tasks increases, using a CPU to schedule tasks becomes inefficient. Cluster scheduling running within the network can...

Full description

Bibliographic Details
Main Authors: Stoyanov, R, Armour, W, Zilberman, N
Format: Conference item
Language:English
Published: Association for Computing Machinery 2022
_version_ 1811140490545856512
author Stoyanov, R
Armour, W
Zilberman, N
author_facet Stoyanov, R
Armour, W
Zilberman, N
author_sort Stoyanov, R
collection OXFORD
description Efficient use of computing clusters is crucial in large-scale data centers: even small gains in utilization can save millions of dollars. However, as the number of microsecond-scale tasks increases, using a CPU to schedule tasks becomes inefficient. Cluster scheduling running within the network can solve this problem, and brings additional benefits in scalability, performance and power efficiency. However, the resource constraints of programmable network devices make network-accelerated cluster scheduling hard. In this paper we propose P4-K8s-Scheduler, a network-accelerated cluster scheduler for Kubernetes implemented on a programmable network device. Preliminary results show that by scheduling Pods in the network at line-rate, P4-K8s-Scheduler can reduce the scheduling overheads by an order of magnitude compared to state-of-the-art Kubernetes schedulers.
first_indexed 2024-03-07T07:29:32Z
format Conference item
id oxford-uuid:dcc2863d-c86d-4adc-8211-0ebf5b91e3dc
institution University of Oxford
language English
last_indexed 2024-09-25T04:22:49Z
publishDate 2022
publisher Association for Computing Machinery
record_format dspace
spelling oxford-uuid:dcc2863d-c86d-4adc-8211-0ebf5b91e3dc2024-08-10T19:33:59ZNetwork-accelerated cluster schedulerConference itemhttp://purl.org/coar/resource_type/c_5794uuid:dcc2863d-c86d-4adc-8211-0ebf5b91e3dcEnglishSymplectic ElementsAssociation for Computing Machinery2022Stoyanov, RArmour, WZilberman, NEfficient use of computing clusters is crucial in large-scale data centers: even small gains in utilization can save millions of dollars. However, as the number of microsecond-scale tasks increases, using a CPU to schedule tasks becomes inefficient. Cluster scheduling running within the network can solve this problem, and brings additional benefits in scalability, performance and power efficiency. However, the resource constraints of programmable network devices make network-accelerated cluster scheduling hard. In this paper we propose P4-K8s-Scheduler, a network-accelerated cluster scheduler for Kubernetes implemented on a programmable network device. Preliminary results show that by scheduling Pods in the network at line-rate, P4-K8s-Scheduler can reduce the scheduling overheads by an order of magnitude compared to state-of-the-art Kubernetes schedulers.
spellingShingle Stoyanov, R
Armour, W
Zilberman, N
Network-accelerated cluster scheduler
title Network-accelerated cluster scheduler
title_full Network-accelerated cluster scheduler
title_fullStr Network-accelerated cluster scheduler
title_full_unstemmed Network-accelerated cluster scheduler
title_short Network-accelerated cluster scheduler
title_sort network accelerated cluster scheduler
work_keys_str_mv AT stoyanovr networkacceleratedclusterscheduler
AT armourw networkacceleratedclusterscheduler
AT zilbermann networkacceleratedclusterscheduler