Allocation of Virtual Machines in Cloud Data Centers—A Survey of Problem Models and Optimization Algorithms

Abstract

Data centers in public, private, and hybrid cloud settings make it possible to provision virtual machines (VMs) with unprecedented flexibility. However, purchasing, operating, and maintaining the underlying physical resources incurs significant monetary costs and also environmental impact. Therefore, cloud providers must optimize the usage of physical resources by a careful allocation of VMs to hosts, continuously balancing between the conflicting requirements on performance and operational costs. In recent years, several algorithms have been proposed for this important optimization problem. Unfortunately, the proposed approaches are hardly comparable because of subtle differences in the used problem models. This paper surveys the used problem formulations and optimization algorithms, highlighting their strengths and limitations, also pointing out the areas that need further research in the future.

NACER: A Network-Aware Cost-Efficient Resource Allocation Method for Processing-Intensive Tasks in Distributed Clouds

“…Most existing approaches to VM allocation use greedy heuristics [7]. In the following, we argue that such methods are not appropriate for the NARAP problem.…”

“…Most previous work use simple greedy algorithms for optimizing VM allocation [7]. NACER is more powerful in several respects.…”

Toward sustainable data centers: a comprehensive energy management strategy

“…maximize the energy productivity of the data center). On one side, management algorithms can decide the placement of VMs in the physical hosts of the data center, as well as the energetic status of those hosts (on(P-state)/suspended(C-state)/off) [56]. Each VM must be allocated with enough resources of each type to fulfill its demand.…”

“…Whereas data center ecosystems offer new opportunities for energy-driven management, they also encompass new challenges that must be considered [44,56,10], such as the distant geographic locations of data centers, which have an impact in the migration of VMs (increasing the cost significantly (and the consumed energy) and frequently causing service level degradations for the affected customers) and in the data exchange among VMs located in different data centers (increasing the communication latency among them); the independent administrative domains involved in the ecosystem, which have frequently conflicting goals and do not generally disclose information about their energy consumption and energy mix, thus increasing the need for third parties to independently assess energy data of data centers and share this information within the ecosystem; and the importance of the prediction accuracy of the input data (e.g. workload, energy price, renewable energy), which depends on the predictability of each data source and the prediction window length, and can be a downgrading factor on the efficiency of the management algorithms.…”

Optimising for energy or robustness? Trade-offs for VM consolidation in virtualized datacenters under uncertainty

“…However, finding an optimal allocation of VMs to support their resource demands on the given set of physical servers in order to minimise e.g. energy consumption is a very hard computational problem and has led to a number of interesting mathematical modelling approaches in recent years [1,11,14,15,17,19,20].…”