Desempenho Conjunto Paas/IaaS em Nuvem Privada de Borda baseada em Código Aberto: Um estudo experimental para aplicações IoT

Abstract

With the growth in the number of IoT platforms being made available by small, medium, and large service providers. One of the most popular environments for the provision of these IoT services is IaaS, where resources are made available for users to host their applications and platforms on these shared infrastructures. Certain types of IoT applications, such as industrial applications, have stringent requirements such as ultra-low latency, high throughput and low response time, etc. It is not possible to use just any IoT platform to meet these requirements. In order to serve these IoT applications, they must be built to meet the demands and one can also improve their performance by performing their installation using cloud paradigms such as Edge and Fog. These paradigms have the characteristics of being low latency and high throughput because they are close to the end clients. The performance of IoT applications can be influenced by the cloud infrastructure being adopted, the way it is instantiated (VM or container), number of requests, consumption of computational resources among other factors. These factors can be broken down into both PaaS and IaaS, and can use them together. Therefore, this paper proposes the evaluation of joint PaaS/IaaS performance metrics in open source private cloud edge for IoT applications. Characterizing the end-to-end metrics, from the physical and cloud infrastructure to the application for the best convenient choice of configuration (cloud infrastructure, IoT platform, number of clients and protocols) based on the desired application requirements, furthermore perform the identification of bottlenecks and limitations. The implementation and evaluation of two open source edge cloud infrastructure (IaaS) platforms is performed, these being OpenStack and StarlingX. StarlingX is adopted as an edge cloud infrastructure platform with hybrid architecture for the coexistence of container and VMs in the same ecosystem. It is subsequently, performed, an experiment with the IoT platform, ThingsBoard, as PaaS on top of StarlingX exploring the VM and container virtualization mechanism; evaluating and characterizing end-to-end relevant performance metrics set.