Energy Efficiency in Latency-Constrained Application Offloading From Mobile Clients to Multiple Virtual Machines

This paper addresses the energy-latency tradeoff in distributed application offloading, in which an energy-limited handset offloads totally or partially an application to one or several virtual machines (VMs) located in remote locations or access points (APs) close to the mobile terminal (MT). One o...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on signal processing 2018-02, Vol.66 (4), p.1065-1079
Main Authors: Lagen, Sandra, Pascual-Iserte, Antonio, Munoz, Olga, Vidal, Josep
Format: Article
Language:English
Subjects:
Application offloading
Battery savings
Cloud computing
Delays
Eficiència energètica
Energia
Energies
Energy conservation
Energy consumption
Energy efficiency
Energy-latency trade-off
Enginyeria de la telecomunicació
Estalvi
Mobile communication
Optimization
Processament del senyal
Robust design
Robustness
Signal processing
Tractament del senyal
Àrees temàtiques de la UPC
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper addresses the energy-latency tradeoff in distributed application offloading, in which an energy-limited handset offloads totally or partially an application to one or several virtual machines (VMs) located in remote locations or access points (APs) close to the mobile terminal (MT). One of the APs (the serving AP) provides radio access to the MT and is connected to the VMs through nonideal backhaul (BH) links. In this setting, we optimize the offloading strategy (including the joint optimization of radio and computational resources) to minimize the energy consumption at the MT subject to a maximum latency constraint. In addition, we propose robust designs to cope with imperfect acquisition of the channel state information (CSI) and the BH parameters. Our findings show that as far as the energy-latency tradeoff is concerned, the optimal order of activation of the VMs does not depend on their processing capabilities but the delays of the BH links. However, once a VM is selected to participate in the processing, the optimal amount of processing allocated to such VM depends on its computational capabilities as well as on the features (capacity and delay) of the BH link. Additionally, offloading decisions become more conservative as the uncertainty in CSI and BH parameters increases.
ISSN:1053-587X
1941-0476
DOI:10.1109/TSP.2017.2778692