Task Offloading Methods in Fog-based IOT

Abstract

Fog computing enables efficient task processing with low processing latency and bandwidth overhead between devices and the cloud in Internet of Things (IoT) networks. The reliability of these networks is critical as fog servers must efficiently allocate and offload increasing types of tasks to appropriate resources. In dynamic networks like Connected Vehicle Networks (CVNs), vehicles communicate with each other and various IoT devices within the transportation infrastructure. In CVNs, vehicles may experience task offload issues due to dynamic conditions. This thesis presents solutions to increase resource utilization and reduce processing latency in task offload in fog computing based IoT networks. Task offload is improved by optimizing resource allocation with P4 technology. In addition, mobility-oriented protocols are developed to manage the dynamic structure in CVNs.

Firstly, a novel task offloading scheme called Task Offloading Scheme with P4 (TOS-P4) is proposed for fog-based IoT networks using p4 technology to improve network performance regarding latency and computation overhead. Evaluated in an Intelligent Transportation System (ITS) scenario, TOS-P4 is compared to a conventional model called Task Offloading Scheme with a Software-Defined Networking Controller (TOS-SDN). Experimental results indicate that TOS-P4 is 6.54 times more efficient than TOS-SDN in the waiting times of tasks received by Resource Poor (RP) fog servers. Additionally, the average waiting time is 30 times longer on RP servers than on Resource Rich (RR) servers. Traditional fixed position fog networks are inadequate for dynamic traffic info capturing, whereas mobile fog computing enables rapid task processing using nearby vehicles. Secondly, this thesis proposes a mobility-driven joint task offloading and resource utilization protocol called MobTORU to optimize resource utilization and efficient task processing in CVNs. The proposed protocol and algorithm are evaluated using a real-world dataset containing actual vehicular mobility traces, demonstrating 93.8% overall system efficiency and 99.9% efficiency in the resource utilization of offloaded tasks. Task offloading in CVNs poses significant challenges due to high computational demands and dynamic network conditions. Traditional static fog networks struggle to adapt to dynamic traffic network conditions, inefficiently allocate resources, and incur higher utilization and maintenance costs. Mobile fog computing provides adaptive and efficient task processing. Lastly, this thesis introduces a novel mobility-driven protocol, Mx-TORU, which combines multi-hop task offloading and resource utilization optimization to improve task processing efficiency in CVNs. Mx-TORU aims to utilize resources efficiently, complementing the previously proposed MobTORU protocol. Extensive experiments demonstrate the effectiveness of our approach, showing that multi-hop task offload methods outperform single-hop methods by up to 17.8% in effective resource utilization rates. Additionally, the proposed protocol and RELiOff algorithm consistently exhibit at least a 5% increase in effectiveness across various test scenarios.

Overall, this thesis addresses task offloading challenges in fog-based IoT networks, especially in CVNs. By using P4-supported task offloading scheme and mobility-driven protocols, it improves resource efficiency and reduces latency. Comparative analyses of thesis shows more superior performance over traditional methods, enhance fog computing's efficiency and reliability in IoT. It contributes new methodologies and the way for future research in optimizing task offloading and resource management in complex networks.