Design of energy‐efficient cloud systems via network and resource virtualization

Abstract

Data centers play a crucial role in the delivery of cloud services by enabling on-demand access to the shared resources such as software, platform and infrastructure. Virtual machine (VM) allocation is one of the challenging tasks in data center management since user requirements, typically expressed as service-level agreements, have to be met with the minimum operational expenditure. Despite their huge processing and storage facilities, data centers are among the major contributors to greenhouse gas emissions of IT services. In this paper, we propose a holistic approach for a large-scale cloud system where the cloud services are provisioned by several data centers interconnected over the backbone network. Leveraging the possibility to virtualize the backbone topology in order to bypass IP routers, which are major power consumers in the core network, we propose a mixed integer linear programming (MILP) formulation for VM placement that aims at minimizing both power consumption at the virtualized backbone network and resource usage inside data centers. Since the general holistic MILP formulation requires heavy and long-running computations, we partition the problem into two sub-problems, namely, intra and inter-data center VM placement. In addition, for the inter-data center VM placement, we also propose a heuristic to solve the virtualized backbone topology reconfiguration computation in reasonable time. We thoroughly assessed the performance of our proposed solution, comparing it with another notable MILP proposal in the literature; collected experimental results show the benefit of the proposed management scheme in terms of power consumption, resource utilization and fairness for medium size data centers.

Energy-Efficient Networking Solutions in Cloud-Based Environments

“…Considering that different areas of administration and ownership exist in multidomain transfer networks, these solution types cannot be deployed in a centralized consistent way. We have identified two primary studies Decision framework Tarutani et al 2012;Heller et al 2010;Thanh et al 2012;Cordeschi et al 2013;Dong et al 2013;Mahadevan et al 2009;Leivadeas et al 2013;Saha et al 2012;Sun et al 2012;Shirayanagi et al 2013;Nguyen Huu et al 2013;Fang et al 2012;Mahadevan et al 2011;McGeer et al 2010;Huang et al 2011;Kuribayashi 2012;Mann et al 2011;Fang et al 2013] Decision framework [Peoples et al 2012;Kantarci et al 2013] 3 9…”

“…Routing/Switching protocols [Peoples et al 2012] 1 10 Device [Kantarci et al 2013] 1 covering large-scale cloud-based environments and providing the Device solution type. They concentrate on the "IP over wavelength-devision multiplexing" (WDM) optical networks as their backbone transfer network.…”

“…Kantarci et al 2013;Ji et al 2012] Yearly distribution of each solution type from December 2008 to November 2013. et al [2013] evaluate their decision framework in both simulations and empirical experiments.…”

Network performance of multiple virtual machine live migration in cloud federations

“…In fact, increasing any of them results in a higher total migration time, and then in a higher resource occupancy period. This is exemplified by Figures 16 and 17 with relation to the maximum number M of VMs to be migrated for each request, and by Figures 18 and 19 in terms of the large VM memory size multiplying factor u, according to the statistical assumptions made at the beginning of this section that led to the average migration time expressed in (16) for the sequential case. The proposed model allows to determine the limits of the VM migration group size and the VM memory size that are compatible with a desired average performance level, given the amount of computing and communication resources available in the cloud federation.…”

“…A holistic solution to the issue of large-scale, distributed cloud system design is proposed in [16]: the authors formalize an optimal VM placement strategy aimed at minimizing both intra-DC resource demand and inter-DC energy consumption. Other related cross-cloud network optimization works include a resilient optical inter-DC infrastructure planning scheme based on dynamic electricity pricing [17] and a framework for joint computing and communication resource allocation targeted at satisfying green service level agreements [18].…”