A Multi-Cloud Orchestration Model Using Kubernetes For Microservices

Abstract

To enhance resilience and to avoid vendor lock-in, modern enterprises are increasingly deploying microservices across multiple cloud providers. But, coordinating workloads across clouds has remained com- plex, even as Kubernetes has emerged as a de facto standard for con- tainer orchestration. This paper proposes a novel multi-cloud orchestra- tion framework for microservices using Kubernetes as a unifying layer. An architecture is presented for a single control plane to manage mi- croservices across AWS, Azure and Google Cloud Platform clusters. The framework is presented, which includes cross-cloud service discovery, net- work connectivity, and intelligent scheduling of microservices to clusters based on latency, cost, and failover policies, respectively. In early 2020, we compare Kubernetes support in major clouds, highlight their strengths and limitations, and evaluate how our approach leverages their features. Benefits in high availability and load bursting are shown through real- world use cases. Pseudocode is provided for the orchestration logic for de- ploying a microservice across clouds and for handling cross-cloud failover. We find that a multi-cloud Kubernetes model can combine the best fea- tures of each provider and, at the same time, avoid their worst pitfalls, to create robust and flexible microservice deployments.