Mikroservisler için Verimli Yerleştirme Alternatiflerinin Türetilmesi

Abstract

Microservice architecture is a popular approach that relies on modular software components that are loosely coupled, operate with a single functionality that can be independently deployed and communicate with each other via a well-defined interface. Thanks to independent software components, the scalability, update and maintenance of the services are easier. However, the deployment of microservice-based applications is more complicated than monolith applications. A monolith application can be deployed at one time on a group of services behind a load balancer. In a microservice-based application consisting of different services that have more than one runtime instance of each service, thousands of services need to be configured and deployed efficiently. While an application system with a small number of microservices can be manually deployed by an expert who knows the system well, a large number of microservices are often used for the applications in practice. Efficient deployment of the increasing number of services to resources manually in the cloud environment becomes both time consuming and makes the efficient use of resources almost impossible. In particular, resource usage and performance of microservice-based applications depend on the efficient deployment of microservices to the available resources. Existing tools such as Kubernetes, Docker, Apache Mesos don’t consider the runtime data of a microservices and deploy them with minimum ability. Also, in literature, it is foreseed to achieve performance in terms of criteria such as CPU workload, I/O, average response time by developing additional approaches or tool support over existing deployment tools. However, the configuration process for deploying microservices is left to a user in these methods. This situation makes it difficult for the user to work in a system consisting of many services, hence, it prevents the deployment of services to limited resources in an efficient manner.

In this thesis, an approach developed using model driven architecture that provides automatic deployment of services to cloud servers is proposed. Furthermore, deployment of microservices at minimum cost has been performed by developing a tool that supports this approach. The approach and the tool support have been tested using an industrial case study. By effectively deploying microservices to resources, it will be possible to use the resources of the cloud environment efficiently and with higher performance. With the proposed automated deployment approach, it is foreseed to enable an infrastructure for the future academic studies, as well as a contribution to industrial applications with the developed microservice deployment tool.