International Journal of Computing Science and Information Technology, 2018, ISSN: 2278-9669, April 2018 (http://ijcsit.org) A Methodical Review Of Virtualization Techniques In Cloud Computing Salini Suresh 1. Manjunatha Rao 2. Seshadripuram Academy of Business Studies, Dr. Ambedkar Institute of Technology, Bangalore, India; Bangalore, India, Received January 2018 Abstract IT enterprises has widely embraced virtualization trend to incorporate autonomic computing. Server virtualization em- powers the creation of virtual environments that are adept to run operating system and applications. Recent IDC survey forecasts that companies can achieve a return of investment of 472 percent by utilizing virtualization. We have present- ed a detailed overview of Hypervisor based and container based implementations. We reviewed the existing approaches of performance evaluation and analysis of the virtualisation tools. Keywords: Virtualization, Hypervisor, VM, Container I. Introduction KVM [4] uses this type of hardware virtualization. Xen [3] uses Para virtualization where hypervisor changes the Virtualization can be defined as “framework or guest in an operating system in contrast to virtual ma- methodology of dividing the resources of a computer chine model, which allows a guest operating system to into multiple implementation environments, by concern- run without modifications. Operating System Level vir- ing one or more notions or technologies such as software tualization or container-based virtualization creates user partitioning, hardware division, time-sharing, partial or instances or containers isolated from each other. complete machine simulation, emulation, worth of ser- Linux-VServer [5], OpenVZ [6] and Linux Containers vice, and many others”[1]. Server virtualization facili- (LXC) [7]) practices container-based virtualization, tates the creation of several isolated virtual instances which is lightweight compared to VM based model. from one physical server. Virtualization conceals the The rest of the manuscript is organised as follows: Sec- identity of server resources from server users and each tion II describes background and related work. virtual server will performance like a physical server. Section III discusses an overview of virtualisation tech- Network virtualization pools the accessible assets of a niques. In Section IV performance analysis of virtualisa- system to the available bandwidth into independent tion techniques is in existing approaches are reviewed. channels that can be allocated to a particular server. Storage virtualization merge physical storage from many II. BACK GROUND AND RELATED devices to a single storage pool. WORK Cloud implementation benefits largely by virtualization Most of the common server operating systems lack con- as it allows the creation of virtual instances of physical figuration isolation, which requires the installation of a resources [2]. The core paybacks of virtualization are separate instance of the operating system for each appli- hardware independence, multitenancy, flexibility, in- cation in a virtual machine. Virtualization techniques creased isolation, and enhanced security. The virtualiza- should provide the resource isolation in a cloud frame- tion systems are categorized into are full virtualization, work [8]. Hypervisor-based systems are the most ac- Para virtualization, and operating system level virtualiza- cepted solution for cloud virtualization [9].Bare metal tion. In Full virtualization operating systems and applica- hypervisor or Type-1 Hypervisors can be mounted on the tions are isolated from the computer hardware by a soft- hardware directly whereas hosted hypervisor or Type-2 ware called Hypervisor thus provides each Virtual ma- hypervisor entails a host operating system. Kernel Virtu- chine (VM) a complete abstraction VMware [3] and al Machine (KVM) is a virtualization tool for Linux [10]. 50 International Journal of Computing Science and Information www.ijcsit.org Special_ issue , April 2018 Copyright © 2018 IJCSIT A guest operating system thus runs on another level users to create and manage applications through above the hypervisor, allowing for true isolation of each APIs[18]. virtual machine. This is the classic VM architecture. An example of this implementation is the VMware ESX C. DOCKER Server and Xen. Linux-VServer [11], OpenVZ [12] Docker is lightweight virtualization technology based on and Linux Containers (LXC) [13] offer container based LXC containers, which encapsulate application and all implementations .While Hypervisors provide abstraction its dependencies in a container, hence with less overhead at the hardware level , containers provide abstraction at than a VM. This Linux based open platform facilitates the application level. Generally, container-based virtual- developers for rapid development and execution of ap- ization systems use Control Groups (Cgroup) [14] to plications, along with the set of tool and workflows for manage resources like CPU, memory. Figures 1(a) and easier deployment and management[19]. 1(b) illustrates Hypervisor based and container-based virtualization. Figure 2 demonstrates an architectural overview of Figure1: Hypervisor based and container-based virtu- Docker. A Docker image basic build component or read-only alization. template to which other components like middleware, web application etc. can be added. Docker registries are III. OVERVIEW OF VIRTUALISA- repositories to share the images [20]. TION TECHNIQUES. In this section, an overview of virtualisation tools are IV. PERFORMANCE ANALYSIS OF given. VIRTUALISATION TOOLS. A. KVM In this section, we reviewed the existing approaches for KVM (Kernel-based Virtual Machine) is a feature of performance evaluation and analysis of the virtualisation Linux that uses full virtualization by], installing an un- tools. modified guest operating system (OS) for Linux process. Sudha et al. [21] have compared the performance of Hy- KVM provisions both emulated I/O devices through pervisor – based system (KVM) and container-based QEMU [15] and paravirtual I/O devices using virtio [16]. system (LXC) and have conducted computation and I/O The full virtualized system provide excellent isolation performance test with Apache Benchmark and IO zone but overhead associated with data transfer is high as cre- File system. They have analysed that LXC containers ation and deployment of disk images for access to stor- out-performed KVM and showed a performance nearer age is time-consuming, leading to wastage of storage to that of the host operating system. space. In antagonistic to the prevalent conviction that David Beserra et al. [22] studies the effect of resource Hypervisor -based systems have high-performance sharing among containers in performance of I/O-bound overhead, studies show that performance overhead of applications in virtualized environments. They observe VM could vary on the job-to-job basis [17]. that physical resources when shared among are shared B. LXC between logically different platforms, LXC and Docker Containerization technology lays its base on LXC (Linux performs as good as non-virtualized environments but containers).Lightweight isolation is usually implemented both the tools do not provide adequate isolation for the by using Kernel Cgroups and namespaces. Namespaces resources. provide resource isolation, network isolation while Miguel G. Xavier et al. [23] investigated that even Cgroups manage the resources, process control and de- though container based system offers near-native per- fine the configuration of a network. LXC has also taken formance, they lack isolation and security. on of Linux Kernel features like chroot, PID and enable 51 International Journal of Computing Science and Information www.ijcsit.org Special_ issue , April 2018 Copyright © 2018 IJCSIT. Jyothi Shetty et al.[24] evaluates overhead of virtualiza- [5] “Linux VServer,” 2012. [Online]. Available: tion with benchmark tests that contain CPU, system, http://linux-vserver.org memory and disk workloads using Phoronix test suite [6] “OpenVZ,” 2012. [Online]. Available: for Docker containers in comparison with VM, and http://www.openvz.org [7] “Linux Containers,” 2012. [Online]. Available: finds Docker outdo VM in terms of CPU and memory http://lxc.sourceforge.net. workloads, the disk I/O performance. Conversely, the [8] An updated performance comparison of virtual machines a hitch of containers is the isolation and security level be- nd Linux containers Wes Felter; Alexandre Ferreira; Ram cause containers share the host kernel. Rajamony; Juan Rubio,2015 IEEE International Sympo- Li et al. [25] evaluated the I/O bandwidth and latency sium on Performance Analysis of Systems and Software reduction in virtual machine hosts and storage servers (ISPASS), 2015,pp. 171 - 172 using real world trace driven simulation. They have pro- [9] X. Xu, H. Yu, and X. Pei, “A novel resource scheduling posed a procedure, Sea-Cache, which fuses approach incontainer based clouds,” in Proc. 17th IEEE de-duplication and a storage server side de-duplication Int. Conf. Comput. Sci.Eng. (CSE 2014). Chengdu, China: IEEE Computer Society, Dec. 2014, pp. 257–264. system that in turn optimize storage server to host data [10] “Linux Containers,” 2012. [Online]. Available: transfers and host side caching. http://lxc.sourceforge.net Sungyong Ahn et al. [26] have examined the perfor- [11] Avi Kivity, Yaniv Kamay, Dor Laor, Uri Lublin, and An- mance of Solid-State Drives (SSDs) in the virtualized thony Liguori.KVM: the Linux virtual