AirEdge: A Dependency-Aware Multi-Task Orchestration in Federated Aerial Computing

Authors: Awada, U., Zhang, J., Chen, S. and Li, S.

Journal: IEEE Transactions on Vehicular Technology

Volume: 71

Issue: 1

Pages: 805-819

eISSN: 1939-9359

ISSN: 0018-9545

DOI: 10.1109/TVT.2021.3127011

Abstract:

Emerging edge computing (EC) systems are currently exploiting attaching portable edge devices on drones for data processing close to the sources, to achieve high performance, fast response times and real-time insights. To this end, existing EC research has proposed several multiple drone-based edge deployments for various purposes, such as data caching, task offloading, real-time video analytics, and computer vision. However, none of them consider the ability of seamlessly integrating edge resources running across multiple drones in a single pool, to holistically manage and control these resources as well as to eliminate vendor lock-in situations. This paper presents an intelligent resource scheduling solution for a federated aerial EC system, called AirEdge, which jointly considers task dependencies, heterogeneous resource demand and drones' flight time. We propose a multi-task execution time estimation and a dispatching policy, to select the closest drone deployment having congruent flight time and resource availability to execute ready tasks at any given time. For the utilization of the drones' attached edge resources, we propose a variant bin-packing optimization approach through gang-scheduling of multi-dependent tasks that co-locates tasks tightly on nodes to fully utilize available resources. Experiments on real-world data-trace from Alibaba cluster trace with information on task dependencies (about 12,207,703 dependencies) and resource demands show the effectiveness, fast executions, and resource efficiency of our approach.

https://eprints.bournemouth.ac.uk/36218/

Source: Scopus

AirEdge: A Dependency-Aware Multi-Task Orchestration in Federated Aerial Computing

Authors: Awada, U., Zhang, J., Chen, S. and Li, S.

Journal: IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY

Volume: 71

Issue: 1

Pages: 805-819

eISSN: 1939-9359

ISSN: 0018-9545

DOI: 10.1109/TVT.2021.3127011

https://eprints.bournemouth.ac.uk/36218/

Source: Web of Science (Lite)

AirEdge: A Dependency-Aware Multi-Task Orchestration in Federated Aerial Computing

Authors: Awada, U., Zhang, J., Chen, S. and Li, S.

Journal: IEEE Transactions on Vehicular Technology

Publisher: IEEE

ISSN: 0018-9545

https://eprints.bournemouth.ac.uk/36218/

Source: Manual

AirEdge: A Dependency-Aware Multi-Task Orchestration in Federated Aerial Computing

Authors: Awada, U., Zhang, J., Chen, S. and Li, S.

Journal: IEEE Transactions on Vehicular Technology

Volume: 71

Issue: 1

Pages: 805-819

ISSN: 0018-9545

Abstract:

Emerging edge computing (EC) systems are currently exploiting attaching portable edge devices on drones for data processing close to the sources, to achieve high performance, fast response times and real-time insights. To this end, existing EC research has proposed several multiple drone-based edge deployments for various purposes, such as data caching, task offloading, real-time video analytics, and computer vision. However, none of them consider the ability of seamlessly integrating edge resources running across multiple drones in a single pool, to holistically manage and control these resources as well as to eliminate vendor lock-in situations. This paper presents an intelligent resource scheduling solution for a federated aerial EC system, called AirEdge, which jointly considers task dependencies, heterogeneous resource demand and drones’ flight time. We propose a multi-task execution time estimation and a dispatching policy, to select the closest drone deployment having congruent flight time and resource availability to execute ready tasks at any given time.

For the utilization of the drones’ attached edge resources, we propose a variant bin-packing optimization approach through gangscheduling of multi-dependent tasks that co-locates tasks tightly on nodes to fully utilize available resources. Experiments on realworld data-trace from Alibaba cluster trace with information on task dependencies (about 12,207,703 dependencies) and resource demands show the effectiveness, fast executions, and resource efficiency of our approach

https://eprints.bournemouth.ac.uk/36218/

Source: BURO EPrints