Virtual machines (VM) are used in cloud computing environments to isolate different software. They also support live migration, and thus dynamic VM consolidation. This possibility can be used to reduce power consumption in the cloud. However, consolidation in cloud environments is limited due to reliance on VMs, mainly due to their memory overhead. For instance, over a 4-month period in a real cloud located in Grenoble (France), we observed that 805 VMs used less than 12% of the CPU (of the active physical machines). This paper presents a solution introducing dynamic software consolidation. Software consolidation makes it possible to dynamically collocate several software applications on the same VM to reduce the number of VMs used. This approach can be combined with VM consolidation which collocates multiple VMs on a reduced number of physical machines. Software consolidation can be used in a private cloud to reduce power consumption, or by a client of a public cloud to reduce the number of VMs used, thus reducing costs. The solution was tested with a cloud hosting JMS messaging and Internet servers. The evaluations were performed using both the SPECjms2007 benchmark and an enterprise LAMP benchmark on both a VMware private cloud and Amazon EC2 public cloud. The results show that our approach can reduce the energy consumed in our private cloud by about 40% and the charge for VMs on Amazon EC2 by about 40.5%.

Part 2: Cloud ComputingInternational audienceNowadays, virtualization is present in almost all computing infrastructures. Thanks to VM migration and server consolidation, virtualization helps reducing power consumption in distributed environments. On another side, Dynamic Voltage and Frequency Scaling (DVFS) allows servers to dynamically modify the processor frequency (according to the CPU load) in order to achieve less energy consumption. We observed that these two techniques have several incompatibilities. For instance, if two virtual machines VM1 and VM2 are running on the same physical host (with their respective allocated credits), VM1 being overloaded and VM2 being underloaded, the host may be globally underloaded leading to a reduction of the processor frequency, which would penalize VM1 even if VM1's owner booked a given CPU capacity. In this paper, we analyze the compatibility of available VM schedulers with DVFS management in virtualized environments, we identify key issues and finally propose a DVFS aware VM scheduler which addresses these issues. We implemented and evaluated our prototype in the Xen virtualized environment

Virtualized cloud infrastructures are becoming very popular as they allow separation of hardware and software management. Infrastructure as a Service (IaaS) is the model providing many advantages to both provider and customer. Minimizing the number of resource (and power consumption) in use is one of the main services that such a cloud model must ensure. This objective can be done either by the customer at the application level (by dynamically sizing the application based on the workload) or by the provider at the virtualization level (by consolidating virtual machines based on the infrastructure's utilization rate). Many research works investigate resource management policies separately at the application level or at the virtualized level. In this paper, we study different strategies for cloud resource management: virtual machine consolidation only, dynamic application sizing only, both policy at the same time (either independent or cooperative). We show that virtual machine consolidation and dynamic application sizing do not fully bring benefits to the cloud provider and customer when being implemented without cooperation. Finally, we propose a cooperative model to improve the efficiency of these strategies, in reducing power consumption and keeping application's Quality of Service. 

In a Cloud computing data center and especially in a IaaS (Infrastructure as a Service), performance predictability is one of the most important challenges. For a given allocated virtual machine (VM) in one IaaS, a client expects his application to perform identically whatever is the hosting physical server or its resource management strategy. However, performance predictability is very difficult to enforce in a heterogeneous hardware environment where machines do not have identical performance characteristics, and even more difficult when machines are internally heterogeneous as for Asymmetric Multicore Processor machines. In this paper, we introduce a VM scheduler extension which takes into account hardware performance heterogeneity of Asymmetric Multicore Processor machines in the cloud. Based on our analysis of the problem, we designed and implemented two solutions: the first weights CPU allocations according to core performance, while the second adapts CPU allocations to reach a given instruction execution rate (Ips) regardless the core types. We demonstrate that such scheduler extensions can enforce predictability with a negligible overhead on application performance.