You are on page 1of 6

Case study of KVM and Xen

Abstract: The cloud computing is the hot topic in this era because of its scalability, flexibility, on-demand, self service and dynamic provisioning features. For the implementation of Cloud the basic thing that need to be installed is Virtual Machine Monitor(VMM)or hypervisor. This paper explains not only need, installation features, overall performance but also comparison of of Hypervisors. Introduction: Cloud computing delivers the Infrastructure as a Service(Iaas),Platform as a Service(PaaS),Software as a Service(SaaS) using virtualization. The virtualization is the concept in which the client is virtually using the physical hardware. Thus the overall system is scalable and multi-tenant. OS virtualization uses multiple guest OS uses simulataneously loaded on host OS. This is supported by the virtual machine monitor called as the hypervisor. The Ubuntu Enterprise Cloud is mainly supported by Eucalyptus architecture which is the hierarchical tree like structure. The Cloud manager is the instructor for all clusters for hand over the task execution. The Cluster controller is used to to search the node which will execute the task. The client on the virtual machine i.e. the instance is not knowing which physical resources he is exactly accessing .That means there should be the component which will schedule the resources from various nodes in or out domain. This component is called as Virtual machine monitor will also schedules the program execution. The virtual machine monitor acts as the traffic cop or supervisor for resources and tasks to complete. This hypervisor allows the hardware virtualization. Thus several instances are facilitated with all resources. Background: That compatibility approach contrasts with that taken by Citrix's virtualization rival VMware, which concentrates its cloud efforts on end-to-end solutions by building services on top of its own hypervisor. For its part, Citrix focuses its enterprise efforts on its free Xen and its involvement in the open-source and open-standards OpenStack cloud. But now the most popular hypervisor is Kernel virtual machine(KVM).

Figure 1.Evolution of hypervisor Architecture:

Figure 2.a shows bare metal hypervisor of the type 1 in which the hypervisor is loaded on physical hardware and on that hypervisor number of guest OS can be installed. While in figure 2.b shows the hosted hypervisor in which the the number of hypervisors are installed on host OS. The bare metal hypervisor is also called as native hypervisor gives the better throughput than hosted one.

The another type is embedded hypervisor in which hypervisor is embedded on the chip which improves the performance. Selection of hypervisor:

1. Overhead of CPU, host and Guest OS memory are evaluated using of different hypervisor. Having less value of CPU usage and memory is selected. Thus virtual machine performance gets enhanced. 2. Each hypervisor is supporting with iSCSI protocol for storage. 3. The selection also matters for homogeneous and heterogeneous network. 4. The hypervisor must schedule for optimum usage of single resource then request for another resources. Scheduling process: The scheduling is performed by use of virtualization concept using server consolidation .The physical resources are virtually used by number of instances .This process is shown in Figure 3 as

Different Types of hypervisors There are various hypervisors are existing with their own functionalities as Table 1:Different types of hypervisors XEN Yes KVM Yes ESXi No Hyper-V No VBox Yes

Open source Type Featues

Para Full Para+full virtualization Virtualization virtualization Slow Fast Fast

Full Full Virtualization Virtualization Fast Recommend for desktop

Hypervisor Features: 1. Security:Hypervisor are mostly built in within the linux version acting as the firewall allocates the separation between the user address space and kernel address space. 2. Memory management: Kernel Same Page Merging is the mechanism in which memory of each virtual machine is scanned and merged into the single copy page to avoid cache miss. It is stored with the single copy i.e.If the guest try to change this shared page ,he is accessing his own private copy. 3. Hardware support: Embedded hypervisors ae built on chips. 4. Storage: Hypervisors are using NAS/SAN support for SCSI 5. Live migration : It is the ability to move running virtual machine between various physical hosts with no interruption to service.

6. Guest OS: In para virtualization the guest OS support is limited for linux versions only while in full virtualization,it is not. 7. Performance and scalability: the hypervisor is giving high response ,throughput and scalable. 8. Resource management and process scheduling: the hypervisor is capable of handling allthese because of its flexible architecture. 9. Lower Latency and higher determination: Requests for cloud are divided and conquered from virtual machine and processed fast. Xen There are number of domains in cloud .Lets take the 1st domain is called as Dom0 and the last domain is called as DomU.The Dom0 has high priority than DomU. Hence if more number of domains gets increases, the distribution or allocation of resources to last domain gets interrupted.Thus the data processing or efficiency of overall system gets decreases.Also If the virus infects any domain, other domains also gets infected .This is the disadvantage of using Xen but pass through model handles it.Figure 4 shows xen pass through model in which Dom0s load gets reduced and the reduction of data transmission between hardware to DomU using device drivers.

Figure 4.Xen Pass through model KVM

In KVM there are three kinds of modes Kernel mode, User mode and Guest mode. Again the guest mode has two types guest-user mode and guest-kernel mode. Non I/O tasks are called as guest-user mode while I/O tasks for data called as guest-kernel mode. These requests are handled by Qemu(Quick Emulator) acting as the emulator in both user mode and computer emulation. Performance analysis: Following table explains the performance comparison beween KVM and Xen Table 2: Performance analysis between KVM and Xen Feature Security KVM 1)Security Enhanced Linux (SELinux) 2)Resource isolation 3) Use of Svirt for security policies 1)Kernel same page Merging 2) swap memory 3)Memory virtualization 4)Less number of guest allowed 1)Live migration Xen 1) Guest isolation 2) Access privilege

Memory Management


Scheduling policy

1)Managed by standard Linux kernel 2)KVM quality of service features are enhanced by using control groups acting as the resource scheduler ,Completely fair scheduler(CFS)acting as the process scheduling algorithm, network name space,

1)Memory over commit strategy i.e. sum of the total memory for VM is greater than total physical memory 2)Maximum number of guests are allowed. 1)Increases with Paravirtualization of CPU,I/O, Memory virtualization 2)Pass through model allows guest domain directly contact with physical hardware 3)Improves the response with less the processing time 4)Ability to process guest OS without affecting Host OS overhead. 1)Various CPU scheduling algorithms such as round robin algorithms, 2) Atropos is the algorithm used to reserve the shares of CPU 3)Preemptive scheduling 4)Scheduling are done by hyper calls having parameters Yield-calling runnable domain Block-Removes the domain from queue

Shutdown-used to turn off the runnable domain

Conclusion This paper show how the hypervisor is useful for process scheduling and resource management. Paper also reveals the different types of hypervisor, features of hypervisor out of that the paper has focused on Xen and KVM. The paper also shows the performance analysis between them. It has been seen that Xen and KVM are both having good response and throughput. Both are their own qualities and features. The developer has to decide which he want to use.