Gerenciamento de redes em um ambiente de computação em névoa

Abstract

This work aims to analyze and evaluate network monitoring and management software in a Fog computing architecture in order to identify which one best suits the requirements of this computing environment. To support this evaluation, concepts of Fog computing, monitoring software and ISO/IEC standards for software evaluation were studied. To generate an understanding of the requirements to be monitored and managed, a survey was carried out in the form of a Systematic Review, in which several of the requirements necessary to manage the control of QoS in Fog environments were raised, as well as the tools to be tested, based on the work of several authors. The CAPES database was used with the keywords: "Fog Computing", "Requirements", "QoS" and "Management". The listed requirements were divided into categories based on similarity, finally, the categories with the highest number of mentions that fit the scenario available for testing were selected. These listed requirements were separated between agents, network and server. The selection of the tools was based on the number of repetitions found in the literature review, compatibility with the Linux operating system and having free code or a trial version available, in addition to having documentation to assist in its use. In this way, the tools were selected: Ganglia, Nagios and Zabbix. Based on these studies, metrics, criteria and test cases were defined for the evaluation. The test environment had a local Fog server, two Fog nodes, an access point and a Cloud server. The Fog server was virtualized through the VirtualBox tool, for the Cloud server, a virtual machine was used on Google Cloud, while for the agents, virtualization was not used. Testing tools were installed on the local server. To simulate a Fog environment and simulate the traffic of a characteristic environment, with high data flow, a Python application was developed divided into three modules, the agent module collects CPU usage data and sends it to the Fog server, which by in turn, calculates the arithmetic average of each agent machine during a period of one minute and then sends these averages to the Cloud server, which stores the information in a database. After the tests were carried out, the tools were evaluated using the ISO/IEC 25010 standard that defines the quality characteristics for a software product. The feature of functionality was chosen in order to assess whether the tools are suitable for the Fog architecture.Within this characteristic, two sub-characteristics were selected: adequacy and interoperability. To evaluate these items, external metrics were adopted, which provide to run the tools and collect data in the execution environment. The results achieved showed that none of the tools is suitable for a Fog environment. The tool that got the best result was Zabbix, followed by Nagios and Ganglia, which got the last place. [resumo fornecido pelo autor]