Efficient latency-and-energy-aware IoT-fog-cloud task orchestration: novel algorithmic approach with enhanced arithmetic optimization and pattern search

Nowadays, the load on datacenters has become more and more due to the unprecedented growth of diversified data from many IoT devices; hence, resource utilization has become more difficult. So, Cloud computing emerged as an appealing solution to leverage IoT devices' computational, memory, capacity and resource constraints. However, the physical gap between IoT devices and the Cloud datacentre resulted in latency overhead and incurred high energy consumption due to the centralized nature of datacenter. Thus, Fog computing is introduced as a complementary to Cloud computing to minimize the physical gap and thus, latency by bringing the computations closer to the end devices. However, the placement of tasks to a particular Fog node (FN)/cloud virtual machine (VM) is an NP-hard problem with much higher computational time. Thus, metaheuristic algorithms are a way out as they have a polynomial time complexity. The arithmetic optimization algorithm (AOA) along with the Pattern search algorithm is one of the optimization algorithms, implemented in order to minimise the makespan and maximise resource utilization. This paper proposes an improved AOA when combined with a pattern search algorithm helps in migrating the tasks from an overloaded FN to an underloaded FN more effectively. Finally, the proposed algorithm is compared with similar optimization algorithms on various parameters. From the simulations, it is evident that the proposed AOA-PS-based task scheduling approach outperformed others with a percentage of improvement of 24.2%, 18.6%, 16%, 5.4%, and 23.4% for makespan, resource utilization, response time, energy consumption, and degree of imbalance, respectively.