A Proactive Cloud Application Auto-Scaler using Reinforcement Learning

Albin Heimerson; Johan Eker; Karl-Erik Årzén

doi:10.1109/UCC56403.2022.00040

A Proactive Cloud Application Auto-Scaler using Reinforcement Learning

Albin Heimerson, Johan Eker, Karl-Erik Årzén

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

78 Downloads (Pure)

Abstract

This work explores the use of reinforcement learning to design a proactive cloud resource auto-scaler that is able to predict usage across a distributed microservice application. The focus is on serving time-sensitive workloads, e.g., industrial automation, connected XR/VR (eXtended Reality/Virtual Reality), etc., where each job has a deadline and there is some cost associated with missing a deadline. A simple workload model, as well as a microservice application model, is presented. A reinforcement learning agent is trained to identify workloads and predict needed utilization for identified service chains. The results are compared to standard purely reactive techniques.

Original language	English
Title of host publication	UCC '22: Proceedings of the 15th IEEE/ACM International Conference on Utility and Cloud Computing
Pages	213-220
Number of pages	8
ISBN (Electronic)	9781665460873
DOIs	https://doi.org/10.1109/UCC56403.2022.00040
Publication status	Published - 2022
Event	15th ACM-IEEE International Conference on Formal Methods and Models for System Design - New York, United States Duration: 2017 Sept 29 → 2017 Oct 2 Conference number: 15th

Conference

Conference	15th ACM-IEEE International Conference on Formal Methods and Models for System Design
Country/Territory	United States
City	New York
Period	2017/09/29 → 2017/10/02

Subject classification (UKÄ)

Control Engineering

Access to Document

10.1109/UCC56403.2022.00040

DML_ICC_2022_ServiceMeshRL_short__No_template_

Autonomous datacenter for long term deployment: AutoDC
Eker, J. (Researcher), Årzén, K.-E. (Researcher) & Heimerson, A. (Researcher)
2018/12/01 → 2021/10/06
Project: Research
Event-Based Control of Stochastic Systems with Application to Server Systems
Cervin, A. (PI), Thelander Andrén, M. (Research student), Bernhardsson, B. (Assistant supervisor), Soltesz, K. (Assistant supervisor), Heimerson, A. (Research student) & Ruuskanen, J. (Research student)
Swedish Research Council
2018/01/01 → 2022/12/31
Project: Research

Cite this

@inproceedings{980385f4ff4b413baf01554c56552369,

title = "A Proactive Cloud Application Auto-Scaler using Reinforcement Learning",

abstract = "This work explores the use of reinforcement learning to design a proactive cloud resource auto-scaler that is able to predict usage across a distributed microservice application. The focus is on serving time-sensitive workloads, e.g., industrial automation, connected XR/VR (eXtended Reality/Virtual Reality), etc., where each job has a deadline and there is some cost associated with missing a deadline. A simple workload model, as well as a microservice application model, is presented. A reinforcement learning agent is trained to identify workloads and predict needed utilization for identified service chains. The results are compared to standard purely reactive techniques.",

author = "Albin Heimerson and Johan Eker and Karl-Erik {\AA}rz{\'e}n",

year = "2022",

doi = "10.1109/UCC56403.2022.00040",

language = "English",

pages = "213--220",

booktitle = "UCC '22: Proceedings of the 15th IEEE/ACM International Conference on Utility and Cloud Computing",

note = "15th ACM-IEEE International Conference on Formal Methods and Models for System Design ; Conference date: 29-09-2017 Through 02-10-2017",

}

Heimerson, A , Eker, J & Årzén, K-E 2022, A Proactive Cloud Application Auto-Scaler using Reinforcement Learning. in UCC '22: Proceedings of the 15th IEEE/ACM International Conference on Utility and Cloud Computing. pp. 213-220, 15th ACM-IEEE International Conference on Formal Methods and Models for System Design, New York, New York, United States, 2017/09/29. https://doi.org/10.1109/UCC56403.2022.00040

TY - GEN

T1 - A Proactive Cloud Application Auto-Scaler using Reinforcement Learning

AU - Heimerson, Albin

AU - Eker, Johan

AU - Årzén, Karl-Erik

N1 - Conference code: 15th

PY - 2022

Y1 - 2022

N2 - This work explores the use of reinforcement learning to design a proactive cloud resource auto-scaler that is able to predict usage across a distributed microservice application. The focus is on serving time-sensitive workloads, e.g., industrial automation, connected XR/VR (eXtended Reality/Virtual Reality), etc., where each job has a deadline and there is some cost associated with missing a deadline. A simple workload model, as well as a microservice application model, is presented. A reinforcement learning agent is trained to identify workloads and predict needed utilization for identified service chains. The results are compared to standard purely reactive techniques.

AB - This work explores the use of reinforcement learning to design a proactive cloud resource auto-scaler that is able to predict usage across a distributed microservice application. The focus is on serving time-sensitive workloads, e.g., industrial automation, connected XR/VR (eXtended Reality/Virtual Reality), etc., where each job has a deadline and there is some cost associated with missing a deadline. A simple workload model, as well as a microservice application model, is presented. A reinforcement learning agent is trained to identify workloads and predict needed utilization for identified service chains. The results are compared to standard purely reactive techniques.

U2 - 10.1109/UCC56403.2022.00040

DO - 10.1109/UCC56403.2022.00040

M3 - Paper in conference proceeding

SP - 213

EP - 220

BT - UCC '22: Proceedings of the 15th IEEE/ACM International Conference on Utility and Cloud Computing

T2 - 15th ACM-IEEE International Conference on Formal Methods and Models for System Design

Y2 - 29 September 2017 through 2 October 2017

ER -

A Proactive Cloud Application Auto-Scaler using Reinforcement Learning

Abstract

Conference

Subject classification (UKÄ)

Access to Document

Fingerprint

Projects

Autonomous datacenter for long term deployment: AutoDC

Event-Based Control of Stochastic Systems with Application to Server Systems

Cite this