Innovative techniques for deployment of microservices in cloud-edge environment

PhD ThesisThe evolution of microservice architecture allows complex applications to be structured into independent modular components (microservices) making them easier to develop and manage. Complemented with containers, microservices can be deployed across any cloud and edge environment. Although containerized microservices are getting popular in industry, less research is available specially in the area of performance characterization and optimized deployment of microservices. Depending on the application type (e.g. web, streaming) and the provided functionalities (e.g. ltering, encryption/decryption, storage), microservices are heterogeneous with speci c functional and Quality of Service (QoS) requirements. Further, cloud and edge environments are also complex with a huge number of cloud providers and edge devices along with their host con gurations. Due to these complexities, nding a suitable deployment solution for microservices becomes challenging. To handle the deployment of microservices in cloud and edge environments, this thesis presents multilateral research towards microservice performance characterization, run-time evaluation and system orchestration. Considering a variety of applications, numerous algorithms and policies have been proposed, implemented and prototyped. The main contributions of this thesis are given below: Characterizes the performance of containerized microservices considering various types of interference in the cloud environment. Proposes and models an orchestrator, SDBO for benchmarking simple webapplication microservices in a multi-cloud environment. SDBO is validated using an e-commerce test web-application. Proposes and models an advanced orchestrator, GeoBench for the deployment of complex web-application microservices in a multi-cloud environment. GeoBench is validated using a geo-distributed test web-application. - i - Proposes and models a run-time deployment framework for distributed streaming application microservices in a hybrid cloud-edge environment. The model is validated using a real-world healthcare analytics use case for human activity recognition.