A new Scheduling Scheme in Fog Computing system using Deep Reinforcement Learning Algorithm

Fog Computing (FC) has recently emerged as a promising new paradigm

that provides resource-intensive Internet of Things (IoT) applications with low

latency services at the network edge. However, the limited capacity of

computing resources in Fog colonies poses great challenges for scheduling and

allocating application tasks. In this dissertation, an Intelligent Scheduling

Strategy Algorithm in a Fog Computing system based on Multi-Objective Deep

Reinforcement Learning (MODRL) is proposed. MODRL algorithm select

nodes (Fog nodes or Cloud nodes) for task processing based on three

objectives; current node’s Load, node Distance, and task Priority. MODRL is

a smart method that integrates the ideas of Multi-Objective Optimization and

Deep Reinforcement Learning to tackle intricate decision-making situations

involving several conflicting objectives. This technique is especially valuable

in situations when there is a requirement to maximize numerous criteria

simultaneously, even if they do not exactly line, and where trade-offs need to

be taken into account. The proposed model addresses two main problems; task

allocation and task scheduling. Employ three Deep Reinforcement Learning

(DRL) agents based on a Deep Q Network (DQN), one for each objective. It is

a specific form of Artificial Neural Network structure employed in

Reinforcement Learning. The DQN algorithm utilizes a Deep Neural Network,

commonly a Convolutional Neural Network (CNN), to estimate the Q-function.

This enables the model to effectively process intricate input domains. However,

this is a more challenging scenario because there is a trade-off among these

objectives, and eventually, each algorithm may select different processing

nodes according to its own objective, which brings to a Pareto Front problem.

To solve this problem, propose using Multi-Objective Optimization, a Non

dominated Sorting Genetic Algorithm (NSGA2), and a Multi-Objective

Evolutionary Algorithm based on Decomposition (MOEA/D), which are Multi-VII

Objective Optimization algorithms that can choose the optimal node by

considering three objectives.

Simulation investigation and experiments using a Python environment

with TensorFlow, PyTorch, Pymoo, and PQDM libraries in PyCharm, which is

a powerful Python IDE, to simulate and train the Intelligent Scheduling

Strategy. As well as, Virtualized data using MatPlotLib in the Jupyter

Notebook, indicates that the proposed Intelligent Scheduling Strategy could

attain better results for the several employed efficiency, adaptability, and

performance metrics: Task Completion Time, Makespan, Transmission Delay,

Queueing Delay, Processing Delay, Propagation Delay, Computational Delay,

Latency, Network Congestion, Throughput, CPU Load, and Storage

Utilization, with an average value of 2.02ms, 10ms, 25ms, 2ms, 1.0ms,

9.5ms,3ms, 3.5ms, 0.10ms, %100, %10, and % 99, respectively.