Adaptive Resource Dimensioning with Joint Workload Placement for Cloud Stack Layers

Distributed cloud combines the best of telco and cloud technology and can run any application across multiple sites. Each site might be equipped with multiple virtualization layers (multi-layer cloud stack), e.g., OpenStack and/or Kubernetes. Workloads can be hosted in different virtualization layers. While this brings flexibility and elasticity for service provisioning, it also increases the complexity of life cycle management (LCM) during service instance design and assign. Especially, the compute and networking resources are typically overprovisioned due to a lack of an automated mechanism to scale them up/down or in/out according to their use. Adaptive resource dimensioning of cloud layers is needed to significantly smooth the process of service provisioning and to better utilize the virtualization layers’ resources. To address these challenges, in this paper, we propose a joint resource dimensioning and workload placement solution with multiple virtualization layers during service instance design and assign time. We demonstrate the feasibility of our solution with illustrative examples.