Efficient Task Scheduling and Resource Allocation Using Osmotic Computing

Abstract

The rapid proliferation of Internet of Things (IoT) devices, projected to reach 30.9 billion by 2025, has led to an unprecedented surge in data generation and a heightened demand for real-time, low-latency processing. As IoT applications become increasingly critical in sectors such as healthcare, transportation, and smart cities, traditional cloud-centric computing models face significant challenges in meeting the stringent requirements for timely data processing and service delivery. The inherent latency and bandwidth limitations of centralized cloud architectures create bottlenecks that hinder their effectiveness in handling the vast and dynamic data streams generated by IoT devices. To address these challenges, there is a pressing need for computing paradigms that can adapt to the distributed nature of modern networks and efficiently manage the high volume of data generated. This work explores Osmotic Computing (OC), a novel paradigm designed to bridge the gap between edge and cloud computing by optimizing resource allocation and service migration across heterogeneous environments. OC leverages emerging technologies such as 5G and Mobile Edge Computing (MEC) to enhance the responsiveness and scalability of computing systems, enabling them to meet the demands of real-time applications. This work focuses on developing an efficient framework for task scheduling and resource allocation in 5G networks, Intelligent Transportation Systems (ITS), and IoT environments.

The first framework, OCTRA-5G, is developed to address the limitations of traditional task scheduling and resource allocation in 5G networks. Traditional models struggle to efficiently manage the increased complexity and demand of 5G environments. OCTRA-5G utilizes OC principles to segregate services into microservices and macroservices, optimizing their scheduling and migration. Through simulations on sets of 10, 20, and 30 gNBs, the framework demonstrated substantial improvements in performance using algorithms such as FCFS, SJF, and PS, effectively reducing latency and improving overall efficiency.

The second framework, OsCoMIT, is introduced to tackle the challenges in Intelligent Transportation Systems (ITS). With the rise in intelligent vehicles, there is a critical need for efficient resource allocation at the edge network to handle service requests swiftly and effectively. OsCoMIT employs a Proportional Fairness (PF) algorithm to manage computational and memory resources for these vehicles. This framework improves upon traditional algorithms like FCFS and PS by enhancing resource utilization and system performance. The results of the framework are validated through statistical analysis using ANOVA which shows that OsCOMIT performed better than other algorithms.

The third framework, DQN-Osmosis, addresses the dynamic nature of IoT, edge, and cloud environments by introducing Deep Q-Networks (DQN) for intelligent decision-making. In rapidly changing network conditions, traditional decision-making approaches may not adapt swiftly enough. DQN-Osmosis uses reinforcement learning to optimize service migration and task offloading, significantly improving resource allocation and processing efficiency compared to Random Agent, Q-Learning, and SARSA. The effectiveness of this approach was confirmed through the Wilcoxon signed-rank test. The fourth framework, μ − osmotic, a novel approach designed to address the challenges of dynamic resource allocation in Intelligent Transportation System. As these devices generate vast amounts of data, the need for efficient computational paradigms at the network edge becomes critical. The proposed μ−osmotic framework leverages Osmotic Computing principles and the Advantage Actor-Critic (A2C) algorithm to optimize the distribution of service requests between edge and cloud resources. By dynamically managing resources on the basis of real-time values of the metrics such as CPU usage, memory consumption, and energy efficiency, μ−osmotic enhances service performance, reduces latency, and ensures effective resource utilization. Through comprehensive evaluation and comparison with other algorithms, the framework demonstrates significant improvements.