kFHCO - Optimal VM Consolidation via k -Factor Horizontal Checkpoint Oversubscription.

Consolidation is one pivotal practice for virtual machine (VM) placement in compute cloud to reduce the cost and energy consumption. Oversubscription is often used by cloud service provider (SP) to fulfill this consolidation goal. But too much consolidation may overload cloud servers, which adversely impact the Quality of Service (QoS) of cloud services. Actions need to be taken by the cloud management platform to maximize consolidation while reducing the risk of overload. There are mainly four approaches to mitigate overload: Ballooning technique (i.e., resource cooperation), VM live migration, optimized placement of VMs and proactive machine learning. But still those works suffer from some disadvantages such as incompatibility for cross-platform cloud services, inability to handle overload radically or being limited to single physical machine (PM) scenario. In this paper, we propose a VM checkpoint placement strategy k-Factor Horizontal Checkpoint Oversubscription (kFHCO) which maximizes consolidation while constraining the chance of overhead. Since the consolidation problem is usually NP-hard like Bin Packing, we first determine the optimal k and target VMs/PMs using a probability-based trigger for the checkpoints. We then formulate the problem as an optimization problem and solve it through an approximation algorithm. We compared our approach with state-of-art solutions using different strategies of VM migration. Simulation results show that our solution can achieve better consolidation level and save cost in various constraints compared with state of the arts.