VM Placement via Resource Brokers in a Cloud Datacenter
Resource management in cloud datacenters is one of the most important issues for cloud service providers because it directly affects their profit. Energy and performance guarantee are two major concern of it. In energy aspect, the total estimated energy b
- PDF / 1,911,646 Bytes
- 27 Pages / 439.37 x 666.142 pts Page_size
- 65 Downloads / 170 Views
VM Placement via Resource Brokers in a Cloud Datacenter
2.1
Introduction
Resource management in cloud datacenters is one of the most important issues for cloud service providers because it directly affects their profit. Energy and performance guarantee are two major concern of it. In energy aspect, the total estimated energy bill of datacenters is $11.5 billion and their energy bills double every five years [1, 2]. Also, in performance guarantee aspect, many researches insist that performance metrics such as throughput and response time should be considered as well as availability in IaaS SLA [3, 4]. If the IaaS SLA with the performance metrics are applied in public CSPs, performance management should be much more delicate to avoid the SLA penalty cost. Especially in VM placement in the cloud, an application quality of service (QoS)-based approach to allocate workload fairly in physical machines (PMs) and a power-based approach to consolidate VMs maximally are two basic goals respectively [5]. To reduce energy consumption in a cloud datacenter, dynamic right sizing (DRS) is a promising technology to dynamically adjust the number of active servers (i.e., servers whose power is switched on) in proportion to the measured user demands [6]. In DRS, energy saving can be achieved by enabling idle compute nodes that do not have any running VM instances to go into low-power mode (i.e., sleep or shut down). In order to maximize the energy efficiency via DRS, one of the primary adaptive resource management strategies is VM consolidation in which running VM instances can be dynamically integrated into the minimal number of compute nodes based on their resource utilization collected by a hypervisor monitoring module [7]. However, it is difficult to efficiently manage cloud resources because cloud users often have heterogeneous resource demands underlying multiple service applications which experience highly variable workloads. Therefore, inconsiderate VM consolidation might lead to undesirable performance degradation due to workload overloading. Although A virtualization technology allows CSPs to get many benefits such as getting high PM resource utilization via server consolidation and © Springer Nature Singapore Pte Ltd. 2017 C.-H. Youn et al., Cloud Broker and Cloudlet for Workflow Scheduling, KAIST Research Series, DOI 10.1007/978-981-10-5071-8_2
47
2 VM Placement via Resource Brokers …
48
elasticity in their resource usage. In virtualized environments, to guarantee performance in VMs sharing a PM, a hypervisor should provide isolation between VMs. However, while security isolation, fault isolation, and environment isolation are well guaranteed, the current virtualization technology does not provide effective performance isolation. Therefore, the VMs make an effect with each other and it causes performance degradation in the VMs. The phenomenon is called as performance interference [8]. Therefore, addressing the conflict between VM consolidation and dispersion is essential for effective cloud resource management. To achiev
Data Loading...
