Article Preview
Top1. Introduction
A community cloud is a model where the unused resources in a physical datacenter are shared using a cloud middleware layer as depicted in Figure 1. These resources are utilized by the participants of the community for their computing needs. Unlike IaaS cloud providers having datacenters at a fixed region, a community cloud is dispersed over vast locations which increases communication latency. Each isolated cloud offers different quality of service, it is important to place a virtual machine in a location where the communication latency will be minimum. In some cases of community cloud, the virtual machine cannot be placed in the same datacenter due to its limited computing resources. The virtual machine has to be hosted in different datacenter which in turn increases latency in VM to VM communication.
Likewise, it also important to place the virtual machine without compromising resource wastage like (CPU, Memory, RAM, and Power) that increases the cost associated with datacenter infrastructure. By utilizing the resource efficiently and by minimizing the power consumption paves a way to achieve a greener datacenter with less CO2 emission. Deploying efficient hardware modules and effective handling of the resources will consecutively reduce power consumption. In the majority of the datacenters, resources are not efficiently utilized, and the average CPU utilization is below 50% (Barroso & Hölzle, 2007). The cause for the resource underutilization is workload fluctuation and legalities to host workload in specific physical servers. Servers produce heat while operating, it is essential to keep the physical components cool. Overheating will reduce the lifetime of the physical components. Every datacenter is employed with a cooling system to dissipate the heat produced by the servers. An equivalent amount of energy will also be utilized for the cooling system. Apart from cost and efficiency, another problem is high CO2 emission. The datacenters contribute 2% of global CO2 emission (Pettey & Gartner, 2007).
Server Virtualization is introduced to improvise resource utilization. Without server virtualization, it is not possible to run multiple operating systems in the same physical server. Virtualization gives the physical server an ability to run multiple operating systems simultaneously in isolation. Virtualization also favors VM Migration, VM Consolidation, and Load Balancing. When a server is over-utilized using shadow copying techniques a virtual machine can be made to run on another server seamlessly. In the recent decade, usage of virtualization has a tremendous demand in Data Centers due to the dynamic support that virtualization offers. It allows creating independent execution which improves the cohesion in the execution process. Appropriate allocation of virtual machines to physical machines reduces the total usage of hardware. When the physical machine is underutilized, VM can be migrated to other machines to save energy.
Virtualization has impacted modern server technologies and improved the sharing of physical resources for the execution of multiple VM requests simultaneously. The benefits include the sharing of physical resources, globally managed physical machines, also allowing the migration of virtual machines from one physical machine to another (Zou et al., 2014). This introduces the concept of fault tolerance, risk management, efficient handling of resources at hazardous environments.
Virtual Machine Placement (VMP): VMP is the process of allocating the virtual machine to an appropriate physical machine in a distributed data center. Given a significant number of physical machines, the process is to allocate the virtual machine requests to the physical machines considering the given objectives. In this paper, three different objectives are considered to place a virtual machine to a suitable physical machine. The most common misconception arises between virtual machine placement and VM migration, whereas VM migration (Majhi, 2014, 2016) is a course of rebalancing the allocated virtual machine to obtain optimal utilization of datacenter.
This paper is organized as follows: Section 2 explains the related work done in VMP. Section 3 explains the VMP problem definition in detail. Section 4 deals with the bat algorithm and multi-objective optimization. Section 5 describes the results in detail and section 6 concludes the paper with future work.