A Design of Resource Allocation Structure for Multi-Tenant Services in Kubernetes Cluster

Nowadays, for multi-tenant cloud, there are plenty of Kubernetes-based open-source projects that do extraordinary in management, isolation multitenancy. These systems allow multiple customers to share the same clusters by dividing them into multiple logical clusters. To satisfy each tenant different workload demand, resource allocation is handled manually by the multi-tenant system administrator. However, this static resource provisioning method cannot adapt to address workload bursting or scaling enough Pod to serve at peak hour and scaling down the number of Pods at least hour. In our paper, we propose a design of dynamic resource allocation in Kubernetes multi-tenancy system to address the missing dynamic resource allocation in the multi-tenant Kubernetes control plane. We start with a controller interacting with the multi-tenant system, which offers dynamic resource allocation, automated scheduling workloads, and policies-based placement. We also present the preliminary results demonstrating the applicability of allocating resources to tenants' applications and services to adapt to workloads changes. Our initial results show fast Pod creation time with the help of a policies-based scheduler and a resources allocator to tenant's applications.