Datacenters now a days are having very large number of virtual machines. At the present scenario the virtual machine placement is a very crucial issue in cloud infrastructure management. The different VM placement algorithms are being used to reduce the resource and power consumption.
Large scale of datacentres with high speed networks handle extensively increased demand by consolidating thousands of servers with other infrastructures like storage and network systems named as cloud computing. Cloud is a business model which provides on-demand computing resources as a service. Virtualization is a key technology of cloud computing which allows the sharing of the physical resources. The hypervisor or virtual machine monitor (VMM) is being used to deploy virtualization. It behaves like an operating system and take all the decisions related to virtual machines. In a single physical machine, a number of virtual machines can be created. In cloud, each datacenter may have a number of hosts. When a request for the VM is raised with the hypervisor, the act of deciding where to place the VM is known as VM placement problem. Finding a suitable host for placing a VM is a very difficult task as it is considered as a NP hard Problem. There are two methods that are used in VM placement, placing a new VM or migrating a VM from the other hosts. Placing a VM from one host to another host is known as Live Migration. There are various issues like server consolidation, load balancing , Quality of service , server failure etc. that comes during the VM migration. Usually the VM migration degrades the system performance but the best way of VM migration is to use an algorithm where the VM migrates without any system degradation.
VM Placement Model
In a cloud environment each data center can have N heterogeneous physical nodes. Each physical node is characterized by the CPU performance measure in MIPS, RAM size and network bandwidth. Local and global managers are being used in it. Global manager which is also known as virtual machine manager reside on the master node and collect information from the local manager for high availability. If local manager fail then all the required information can be obtained from the global manager. It divides all available hosts of the data center into the six clusters according to the remaining capacity of the resources. When global manager receive resource request from the user it sends it to the appropriate cluster according to the resource requirement of the VM. Each cluster has its own local manager, which continuously receives information from all available hosts in the data center. Based on this information the local manager decides that a particular VM is to be placed on which server. If a local manager fails then the global manager selects the lightest host in the cluster, migrates all VM from that host and makes it local manager for that cluster.
Classification Of VM Placement Algorithms
The result of the VM placement can either saving energy by closing down some severs or it can be maximizing the utilization of the resources. Based on these goal placement algorithm are classified into two type.]
- Power Based approach
- Application QoS based approach
The Main aim of the power based approach is to save the energy. In the following approach VM maps the physical machines that each servers can utilize as their maximum efficiency and the other servers can be shut down depending on load conditions. In the Application QoS based approach a VM maps to the PM with the aim to maximize the QoS delivered by the service provider by using certain Algorithms.
Algorithm To Be Used
VM placement problem is a non-deterministic problem. A number of virtual machine placement algorithms have been studied in that run under cloud computing environment.
The efficient algorithm which is used to migrate VM without system degradation is the best fit approach.
Best Fit –It is a greedy approach. In this approach scheduler visits the PM sequentially one by one and search for the enough resources. Each time when the new VM arrives, the scheduler starts searching the PM sequentially in the data center till it finds the best of all PM that has enough resources amongst all. If any of the physical machine does not satisfy the resource requirement of the VM, then a new PM is activated and assigns VM to the newly activated PM.
Virtual machine placement is a technique which is used to map the VM’s to the appropriate (Physical Machine) PM. As the size of the data centers is very large in the cloud computing environment, selecting a proper host for the placement of VM is a very challenging task during the virtual machine migration. I have tried Best-Fit approach which is quite effective in live migration. Every placement algorithm performs well under some specific conditions. So it is a critical task to choose a technique that is suitable for both the cloud user and cloud provider.