Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • apa.csl
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Graph Attention Networks and Deep Q-Learning for Service Mesh Optimization: A Digital Twinning Approach
Karlstad University, Faculty of Health, Science and Technology (starting 2013), Department of Mathematics and Computer Science (from 2013).ORCID iD: 0000-0002-4825-8831
Queen's University Belfast, UK.ORCID iD: 0000-0001-9194-010X
Deggendorf Institute of Technology, Germany.ORCID iD: 0000-0002-9446-8143
KTH, Sweden.
2024 (English)In: Proceedings- IEEE International Conference on Communications / [ed] Valenti M., Reed D., Torres M., Institute of Electrical and Electronics Engineers (IEEE), 2024, p. 2913-2918Conference paper, Published paper (Refereed)
Abstract [en]

In the realm of cloud native environments, Ku-bernetes has emerged as the de facto orchestration system for containers, and the service mesh architecture, with its interconnected microservices, has become increasingly prominent. Efficient scheduling and resource allocation for these microservices play a pivotal role in achieving high performance and maintaining system reliability. In this paper, we introduce a novel approach for container scheduling within Kubernetes clusters, leveraging Graph Attention Networks (GATs) for representation learning. Our proposed method captures the intricate dependencies among containers and services by constructing a representation graph. The deep Q-learning algorithm is then employed to optimize scheduling decisions, focusing on container-to-node placements, CPU request-response allocation, and adherence to node affinity and anti-affinity rules. Our experiments demonstrate that our GATs-based method outperforms traditional scheduling strategies, leading to enhanced resource utilization, reduced service latency, and improved overall system throughput. The insights gleaned from this study pave the way for a new frontier in cloud native performance optimization and offer tangible benefits to industries adopting microservice-based architectures.

Place, publisher, year, edition, pages
Institute of Electrical and Electronics Engineers (IEEE), 2024. p. 2913-2918
Keywords [en]
component, formatting, insert, style, styling
National Category
Computer Sciences Computer Systems
Research subject
Computer Science
Identifiers
URN: urn:nbn:se:kau:diva-97430DOI: 10.1109/ICC51166.2024.10622616Scopus ID: 2-s2.0-85202817543ISBN: 978-1-7281-9055-6 (print)ISBN: 978-1-7281-9054-9 (electronic)OAI: oai:DiVA.org:kau-97430DiVA, id: diva2:1813160
Conference
IEEE International Conference on Communications (ICC), Denver, USA, June 9-13, 2024.
Note

This article was included as a manuscript in the doctoral thesis entitled "Unchaining Microservice Chains: Machine Learning Driven Optimization in Cloud Native Systems" KUS 2023:35.

Available from: 2023-11-20 Created: 2023-11-20 Last updated: 2024-10-07Bibliographically approved
In thesis
1. Unchaining Microservice Chains: Machine Learning Driven Optimization in Cloud Native Systems
Open this publication in new window or tab >>Unchaining Microservice Chains: Machine Learning Driven Optimization in Cloud Native Systems
2023 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

As the cloud native landscape flourishes, microservices emerge as a central pillar for contemporary software development, enabling agility, resilience, and scalability in modern computing environments. While these modular services promise opportunities, particularly in the transformative ecosystem of 5G and beyond, they also introduce a myriad of complexities. Notably, the migration from hardware-centric to software-defined environments, culminating in Virtual Network Functions (VNF), has facilitated dynamic deployments across cloud data centers. In this transition, VNFs are often deployed within cloud native environments as independent services, mirroring the microservices model. However, the advantage of flexibility in cloud native systems is shadowed by bottlenecks in computational resource allocation, sub-optimal service chain placements, and the perpetual quest for performance enhancement. Addressing these concerns is not just pivotal but indispensable for harnessing the true potential of microservice chains.

In this thesis, the inherent challenges presented by cloud native microservice chains are addressed through the development and application of various tools and methodologies. The NFV-Inspector is introduced as a foundational tool, employing a systematic approach to profile and analyze Virtual Network Functions, subsequently extracting essential system KPIs essential for further modeling. Subsequent research introduced a Machine Learning (ML) based SLA-Aware resource recommendation system for cloud native functions. This system leveraged regression modeling techniques to correlate key performance metrics. Following this, PerfSim is proposed as a performance simulation tool designed specifically for cloud native computing environments, aiming to improve the accuracy of microservice chain simulations. Further research is conducted on Service Function Chain (SFC) Placement, emphasizing the equilibrium between cost-efficiency and latency optimization. The thesis concludes by integrating Deep Learning (DL) techniques for service chain optimization, employing both Graph Attention Networks (GAT) and Deep Q-Learning (DQN), highlighting the intersection of DL techniques and SFC performance optimization.

Abstract [en]

In the dynamic cloud native landscape, microservices stand out as pivotal for modern software development, enhancing agility, resilience, and scalability. These services, crucial in the transformative 5G era, introduce complexities such as resource allocation, service chain placement, and performance optimization challenges. This thesis delves into these challenges, emphasizing the development and application of tools and methodologies specific to microservice chains.

Key contributions include the NFV-Inspector, which, while focusing on Virtual Network Functions, is instrumental in profiling and analyzing microservices, extracting vital KPIs for advanced modeling. Further, a Machine Learning-based SLA-Aware system is introduced for resource recommendation in cloud-native functions, utilizing regression modeling to link performance metrics. PerfSim, another simulation framework, is proposed for simulating microservice chains in cloud environments. The thesis also explores Service Function Chain (SFC) placement, aiming to balance cost-efficiency with latency optimization. The thesis concludes by integrating Deep Learning (DL) for service chain optimization, employing both Graph Attention Networks (GAT) and Deep Q-Learning (DQN), showcasing the potentials of DL in SFC optimization.

Place, publisher, year, edition, pages
Karlstad: Karlstads universitet, 2023. p. 36
Series
Karlstad University Studies, ISSN 1403-8099 ; 2023:35
Keywords
Cloud Native Computing, Service Mesh, Performance Modelling, Performance Optimization, Performance Simulation, Machine Learning, Resource Allocation
National Category
Computer Sciences
Research subject
Computer Science
Identifiers
urn:nbn:se:kau:diva-97377 (URN)978-91-7867-420-6 (ISBN)978-91-7867-421-3 (ISBN)
Public defence
2024-01-17, 1B309, Sjöströmsalen, Universitetsgatan 2, Karlstad, 08:30 (English)
Opponent
Supervisors
Available from: 2023-12-04 Created: 2023-11-14 Last updated: 2024-02-02Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopus

Authority records

Gokan Khan, MichelTaheri, JavidKassler, AndreasBoodaghian Asl, Arsineh

Search in DiVA

By author/editor
Gokan Khan, MichelTaheri, JavidKassler, AndreasBoodaghian Asl, Arsineh
By organisation
Department of Mathematics and Computer Science (from 2013)
Computer SciencesComputer Systems

Search outside of DiVA

GoogleGoogle Scholar

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 233 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • apa.csl
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf