Elastic Scheduling for Scaling Virtual Clusters in Cloud Data Center Networks

Abstract

Data center networks (DCNs) have become more extensively applied in cloud computing in recent years. One important mission for DCNs is satisfying the fluctuation of on-demand resources for tenants. During the scaling of virtual clusters (VCs), existing works fail to fully consider placement techniques and the elasticity of the physical resource in the DCN at the same time. To address this, we use elasticity to measure the scaling potential of VCs in terms of both computation and communication resources. In this paper, we consider elastic scaling for existing VCs to maximize elasticity with the constraint of communication cost in the DCN. We achieve this through a resource allocation scheme, VCS, which comes with provable optimality guarantees for single VC scaling. After that, we extend our scheme for the scaling of multiple VCs, and we prove that scaling multiple VCs for the over-time elasticity maximization problem is NP-hard. Based on that, we present the multiple virtual cluster scaling (MVCS) algorithm for offline multiple VC scaling, which can maximize over-time elasticity during a stable time period. Furthermore, we propose two heuristic algorithms, synchronous online MVCS and asynchronous online MVCS, using Bayesian parameter estimation to solve an online scaling with both synchronous and asynchronous incoming rates. Extensive simulations demonstrate that our elastic VC scaling placement schemes outperform the existing state-of-the-art methods in terms of flexibility in the DCN.

Keywords

• Data center networks (DCNs)
• elastic scaling
• virtual cluster (VC)
• optimization
• virtual machine (VM) placement