Article Preview
Top1. Introduction
Information Technology (IT) infrastructures have continued to grow rapidly over the past few years to meet the demand of computational power of modern computer-intensive business and scientific applications. These computing infrastructures consume large amount of electricity which results in operating costs that exceed infrastructure costs. Large electricity-consumption reduce system reliability and durability due to the overheating problem except for overwhelming operating costs. Similarly, the problem of substantial CO2 emissions which help out to the greenhouse effect (Brown, R. 2008). In 2010 the overall power consumption of the data centers in the world was calculated at about 1.5%, which rose to 3% in 2016. In 2013, 260 million watts of electricity consumed by Google data centers which is quite enough to steadily power of 200,000 homes. On the other hand, 30% of cloud data centers average resource usage capacity is 10-15%.
Virtualization technology is one way of reducing a data center’s power consumption. This technology uses consolidation technique for servers and Virtual Machines (VMs). The consolidation allows several VMs into one host and reducing the amount of physical hardware usage and minimizing the resource wastage. Cloud computing patterning recently developed advantage is virtualization and provides on-demand resources on a pay-as-you-go basis over the internet. So, the IT companies to drop using their computing environment’s maintenance costs and the computing needs could be outsourced to the cloud. For the customers who negotiate in terms of Service Level Agreements (SLA), e.g., throughput, response time, it is great significance for cloud providers to offer Quality of Services (QoS) (Buyya, R., et al., 2008). Cloud providers (e.g., Amazon EC2) must deal with power-performance trade-offs to ensure efficient resource management and provide higher resource utilization, as aggressive consolidation of VMs can lead to loss of performance.
Consolidation refers to live migration, which is the process of moving a running VM from one physical server to another. The main goal of live VM migration is to do migration without down-time and move the VMs to minimum number of hosts and switch off the idle servers to power-saving mode. VM consolidation experiences a lot of issues due to performance deprivation and this could be avoided by optimum usage of the resources. Different parameters such as data centers VM Selection, VM placement, host CPU and memory, SLA, and power consumption must be considered when it comes to consolidating and optimizing VM. There are two types of VM consolidation. The first type is known as static VM consolidation where VM size is set up in a single deed using the peak load demand of the workload. VMs are placed in the same host during their entire lifetime. Setting the VM size for the peak load demand confirms that the VM will not be overloaded. However, since the workloads can present variable demand patterns, it can lead to the idleness of the host. The second type is known as dynamic consolidation, where periodical changes in the workload demand are carried out in each VM and based on that the required configuration changes are performed.
Dynamic consolidation is performed in two necessary steps. One is to migrate VMs from underutilized hosts and put them into sleep mode to minimize the number of active hosts. Another step is migrating VMs from overloaded hosts to avoid performance degradation, which may lead to SLA violation of the quality of service requirements. Furthermore, live migration is the way to achieve energy efficiency. The main advantage of live migration is the ability to transfer VM between the hosts with a near to zero downtime. In the real world, the computation demand is very dynamic, and that is why the decision depends on several criteria.