REACT: A Solidarity-based Elastic Service Resource Reallocation Strategy for Multi-access Edge Computing

Medeiros, Alisson; Braun, Torsten; Di Maio, Antonio; Neto, Augusto (2021). REACT: A Solidarity-based Elastic Service Resource Reallocation Strategy for Multi-access Edge Computing. Physical communications, 47, p. 101380. Elsevier 10.1016/j.phycom.2021.101380

[img]
Preview
Text
REACT_FINAL.pdf - Accepted Version
Available under License Publisher holds Copyright.

Download (2MB) | Preview
[img]
Preview
Text
1-s2.0-S1874490721001178-main.pdf - Published Version
Available under License Creative Commons: Attribution (CC-BY).

Download (1MB) | Preview

The Multi-access Edge Computing (MEC) paradigm promises to enhance network flexibility and scalability through resource virtualization. MEC allows telecom operators to fulfill the stringent and heterogeneous requirements of 5G applications via service deployment at the edge of the mobile network. However, current solutions to support MEC struggle to provide resource elasticity since MEC infrastructures have limited resources. The coexistence of many heterogeneous services on the distributed MEC infrastructure makes the resource scarcity problem even more challenging than it already is in traditional networks. Services need distinct resource provisioning patterns due to their diverse requirements, and we may not assume an extensive MEC infrastructure that can accommodate an arbitrary number of services. To address these aspects, we present REACT: a MEC-suppoRted sElfadaptive elAstiCiTy mechanism that leverages resource provisioning among different services running on a shared MEC environment. REACT adopts an adaptive and solidarity-based strategy to redistribute resources from over-provisioned services to under-provisioned services in MEC environments. REACT is an alternative strategy to avoid service migration due to resource scarcity. Real testbed results show that REACT outperforms Kubernetes’ elasticity strategy by accomplishing up to 18.88% more elasticity events, reducing service outages by up to 95.1%, reducing elasticity attempts by up to 95.36%, and reducing over-provisioned resources by up to 33.88%, 38.41%, and 73% for CPU cycles, RAM and bandwidth resources, respectively. Finally, REACT reduces response time by up to 15.5%.

Item Type:

Journal Article (Original Article)

Division/Institute:

08 Faculty of Science > Institute of Computer Science (INF) > Communication and Distributed Systems (CDS)
08 Faculty of Science > Institute of Computer Science (INF)

UniBE Contributor:

Medeiros de Lima, Alisson Patrick; Braun, Torsten and Di Maio, Antonio

Subjects:

000 Computer science, knowledge & systems
500 Science
500 Science > 510 Mathematics

ISSN:

1874-4907

Publisher:

Elsevier

Funders:

[185] University of Bern: Seal of Excellence Fund

Language:

English

Submitter:

Dimitrios Xenakis

Date Deposited:

18 Jun 2021 15:34

Last Modified:

20 Jun 2021 03:10

Publisher DOI:

10.1016/j.phycom.2021.101380

Uncontrolled Keywords:

5G; Multi-access Edge Computing; Resource elasticity; Auto-scaling Self-adaptive; Virtualization

BORIS DOI:

10.48350/156487

URI:

https://boris.unibe.ch/id/eprint/156487

Actions (login required)

Edit item Edit item
Provide Feedback