The revolution of IoT and its capabilities to serve various fields led to generating a large amount of data for processing. Tasks that require an instant response, especially with sensitive delay tasks send to the fog node due to the close distance, and the complex tasks transfer to the cloud data center for its huge computation and storage. However, sending tasks to the fog decreases the transmission delay. Still, it increases the energy consumption of the end users, while transferring tasks to the cloud reduces users' energy consumption but increases the transmission delay due to the long distance; besides, assigning tasks to appropriate resources compatible with task requirements. These are the main challenges in cloudfog computing that need to improve. Thus, this study proposed a Multi-Objectives Grey Wolf Optimizer (MGWO) algorithm to reduce the QoS objectives delay and energy consumption and held in the fog broker, which plays an essential role in distributing tasks. The simulation result verifies the effectiveness of the MGWO algorithm compared to the state-of-the-art algorithms in reducing delay and Energy consumption.INDEX TERMS Cloud-fog computing, delay, energy consumption, grey wolf optimizer, Internet of Things, meta-heuristic, task scheduling.

The Internet of Things (IoT) generates massive data from smart devices that demand responses from cloud servers. However, sending tasks to the cloud reduces the power consumed by the users' devices, but increases the transmission delay of the tasks. In contrast, sending tasks to the fog server reduces the transmission delay due to the shorter distance between the user and the server. However, this occurs at the user end's expense of higher energy consumption. Thus, this study proposes a mathematical framework for workload allocation to model the power consumption and delay functions for both fog and clouds. After that, a Modified Least Laxity First (MLLF) algorithm was proposed to reduce the maximum delay threshold. Furthermore, a new multi-objective approach, namely the Non-dominated Particle Swarm Optimization (NPSO), is proposed to reduce energy consumption and delay compared to the state-of-theart algorithms. The simulation results show that NPSO outperforms the state-of-the-art algorithm in reducing energy consumption, while NGSA-II proves its effectiveness in reducing transmission delay compared to the other algorithms in the experimental simulation. In addition, the MLLF algorithm reduces the maximum delay threshold by approximately 11% compared with other related algorithms. Moreover, the results prove that metaheuristics are more appropriate for distributed computing.

Cloud computing is a ubiquitous platform that offers a wide range of online services to clients including but not limited to information and software over the Internet. It is an essential role of cloud computing to enable sharing of resources on-demand over the network including servers, applications, storage, services, and database to the end-users who are remotely connected to the network. Task scheduling is one of the significant function in the cloud computing environment which plays a vital role to sustain the performance of the system and improve its efficiency. Task scheduling is considered as an NP-complete problem in many contexts, however, the heterogeneity of resources in the cloud environment negatively influence on the job scheduling process. Furthermore, on the other side, the heuristic algorithms have satisfying performance but unable to achieve the desired level of the efficiency for optimizing the scheduling in a cloud environment. Thus, this paper aims at evaluating the effectiveness of the hybrid meta-heuristic that incorporate genetic algorithm along with DE algorithm (GA-DE) in terms of make-span. In addition, the paper also intends to enhance the performance of the task scheduling in the heterogeneous cloud environment exploiting the scientific workflows (Cybershake, Montage, and Epigenomics). The proposed algorithm (GA-DE) has been compared against three heuristic algorithms, namely: HEFT-Upward Rank, HEFT -Downward Rank, and HEFT -Level Rank. The simulation results prove that the proposed algorithm (GA-DE) outperforms the other existing algorithms in all cases in terms of make-span.

No abstract