Article Preview
TopIntroduction
Fog computing has become popular in the last two decades as it provided cloud-like services at the edge of the network to reduce latency (Jalasri & Lakshmanan, 2018). It acts as intermediate node between cloud and IoT Devices to provide computational services to smart applications. There are two types of fog nodes: communication devices like routers, gateways and wireless access points (WAP) (Yousefpour et al., 2019); and computational like cloudlets, which can be thought of as a small-scale version of cloud data centres, providing limited computational resource to IoT devices. Most of the smart applications such as healthcare, military and transportation, integrate IoT and Fog computing technology to improve performance and reduce the maintenance cost (Yousefpour et al., 2019). As the number of IoT device is increasing, the performance of fog computing services is reduced. The increased energy consumption, response time, and performance degradation are significant challenges when IoT devices are increased. Further, there is a significant tradeoff between Virtual Machine utilization and managing the resources. Also a special focus is required in the resource management activities as tasks change their demands for resources dynamically.
Resource management in such a dynamic and complex environment is a challenging task. The efficient resource management framework must satisfy the following requirements (Ghobaei-Arani et al., 2020): i) Consume less energy, ii) reduce the complexity, iii) utilize the resources effectively, and iv) consume less CPU and memory consumption. To fulfil these requirements monitoring system plays an important role in providing accurate information to resource management. Many research algorithms such as FCFS, SJF and Round Robin have been developed for task scheduling and resource management in distributed computing environment but they suffer from underutilization of resource and may not fit for fog computing environment. Hence, there is a need to develop an efficient resource management and monitoring model in fog computing infrastructure.
Figure 1 shows the scenario of proposed resource management and monitoring model in smart industries. The industry contains different manufacturing units that are geographically distributed. Each unit have installed different smart machines and these machines requires computational resource to execute tasks. There are three layers: cloud layer, fog layer and IoT layer. The IoT layer contains various smart machinery with a limited amount of processing capabilities. Smart machines/devices need computational resources to execute the tasks that are generated as per user requirements. If the task needs a high computational resource, then tasks are sent to fog computing. The resource manager in the fog node check the requirements of tasks and allocate appropriate (least loaded) virtual machine if available; otherwise, the task is offloaded to the cloud.
Figure 1. Resource management and monitoring architecture for fog computing environment
In this paper, a resource management and monitoring framework is designed for fog computing environment. The main objectives of proposed model is to energy consumption, CPU and memory consumption and improves the resource utilization rate. The proposed model works in two manifolds i) resource monitoring and ii) resource management. Most of the researchers develop resource management models by assuming efficient resource monitoring is already available. The performance of resource management completely depend on an efficient resource monitoring system that provides accurate information in a timely manner (Birje & Manvi, 2011 & Birje & Bulla, 2019). The resource monitoring model collects the data, analysis and alert the users when the performance is reduced. The resource management model develops bio-inspired algorithms to schedule and allocate resources in fog computing environment. Particle swarm optimization (Potu et al., 2021) algorithms and cat swarm optimization algorithms are used for resource scheduling and allocation. The conventional PSO and CSO consume more time to search optimal solution. To reduce this search time, the average inertia weight update mechanism is designed.
To fulfil above said objectives, following contributions are made