Abstract

The OS load balancing algorithm governs the performance gains provided by a multiprocessor computer system. The Linux's Completely Fair Scheduler (CFS) scheduler tracks process loads by average CPU utilization to balance workload between processor cores. That approach maximizes the utilization of processing time but overlooks the contention for lower-level hardware resources. In servers running compute-intensive workloads, an imbalanced need for limited computing resources hinders execution performance. This paper solves the above problem using a machine learning (ML)-based resource-aware load balancer. We describe (1) low-overhead methods for collecting training data; (2) an ML model based on a multi-layer perceptron model that imitates the CFS load balancer based on the collected training data; and (3) an in-kernel implementation of inference on the model. Our experiments demonstrate that the proposed model has an accuracy of 99% in making migration decisions and while only increasing the latency by 1.9 μs.

Original languageEnglish (US)
Title of host publicationAPSys 2020 - Proceedings of the 2020 ACM SIGOPS Asia-Pacific Workshop on Systems
PublisherAssociation for Computing Machinery
Pages67-74
Number of pages8
ISBN (Electronic)9781450380690
DOIs
StatePublished - Aug 24 2020
Event11th ACM SIGOPS Asia-Pacific Workshop on Systems, APSys 2020 - Tsukuba, Virtual, Japan
Duration: Aug 24 2020Aug 25 2020

Publication series

NameAPSys 2020 - Proceedings of the 2020 ACM SIGOPS Asia-Pacific Workshop on Systems

Conference

Conference11th ACM SIGOPS Asia-Pacific Workshop on Systems, APSys 2020
CountryJapan
CityTsukuba, Virtual
Period8/24/208/25/20

Keywords

  • Linux kernel
  • completely fair scheduler
  • load balancing
  • machine learning
  • neural network
  • operating system

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Hardware and Architecture

Fingerprint Dive into the research topics of 'Machine learning for load balancing in the Linux kernel'. Together they form a unique fingerprint.

Cite this