International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com September 2014, Volume 2 Issue 4, ISSN 2349-4476

Migration of Different Virtual Machines over Cloud Environment with respect to Resource Allocation Prajakta Ugale, S.S.Rahul, Sarita Sangawar , Souvik Saha, Yousuf Syed. Virtual machines and its technology are presently increasingly becoming popular and too much valuable in cloud computing environments due to the advantages of server consolidation, resource isolation and live migration. Live migration of virtual machines can be very effectively used to implement energy saving and load balancing in cloud data centre. However, according to our knowledge, maximum number of the previous research were carried on the implementation of migration technology itself and did not concentrate enough on the impact of resource reservation strategy on migration efficiency, which is relevant to this paper. This paper mainly concentrates on the live migration strategy of different virtual machines with different resource reservation methods with efficiency at the centre stage. Index terms— Virtual machine; Live migration; Resource Reservation; Load balancing; Efficiency; Performance I. INTRODUCTION

Cloud computing is presently receiving significant attention in both academic as well as industrial community as a new computing paradigm to provide dynamically scalable and virtualized resource as a service across the web. By this means, users will now be able to access the multiple resources, such as applications and data, from the cloud anywhere and anytime on demand. Presently, a large number of companies, such as Amazon, Google, Yahoo!, Microsoft, IBM, and Sun are developing their very own cloud computing platforms for commercial consumers and enterprises to access the cloud resources from anywhere through services. Recently, with the rapid development of virtualizationTechnology, more and more data centres use thistechnology to build new generation data centre to support cloud computing due to the benefits such as server consolidation, live migration, and resource isolation. Live migration of virtual machines means the virtual machine seems to be responsive all the time during the migration process from the clients’ perspective. Compared with traditional suspend/resume migration, live migration holds many benefits such as energy saving, load balancing, and online maintenance. Many live migration methods are proposed to improve the migration efficiency. As the live migration technology widely used in modern cloud computing

116

data centre, live migration of multiple virtual machines becomes more and more frequent. Different from the single virtual machine migration, the live migration of multiple virtual machines faces many new problems, such as migration failures due to the insufficient resources in target machine, migration conflicts due to the concurrent migrations, and the migration thrashing due to the dynamic changes of virtual machine workloads. All the above issues should be overcome to maximize the migration efficiency in virtualized cloud data centre environments. Migrating an entire OS and all of its applications as one unit allows us to avoid many of the difficulties faced by process-level migration approaches. Overall, live OS migration is an extremely powerful tool for cluster administrators, allowing separation of hardware and software considerations, and consolidating clustered hardware into a single coherent management domain. Firstly, as we are targeting the migration of active OSes hosting live services, it is critically important to minimize the downtime during which services are entirely unavailable. Secondly, we must consider the total migration time, during which state on both machines is synchronized and which hence may affect reliability. Furthermore we must ensure that migration does not unnecessarily disrupt active services through resource contention (e.g., CPU, network bandwidth) with the migrating OS.

Prajakta Ugale, S.S.Rahul, Sarita Sangawar , Souvik Saha, Yousuf Syed.

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com September 2014, Volume 2 Issue 4, ISSN 2349-4476

We achieve this by using a pre-copy approach in which pages of memory are iteratively copied from the source machine to the destination host, all without ever stopping the execution of the virtual machine being migrated. Pagelevel protection hardware is used to ensure a consistent snapshot is transferred, and a rate-adaptive algorithm is used to control the impact of migration traffic on running services. The final phase pauses the virtual machine, copies any remaining pages to the destination, and resumes execution there. I. PROBLEM ANALYSIS AND DEFINITION

The goal of our system is to migrate the wholesystem state of a VM from the source to the destination machine, including its CPU state, memory data, and local disk storage data. During the migration time the VM keeps running. This section describes the key metrics and requirements for a whole system live migration. A. Definition of the Metrics

The following metrics are usually used to measure the effectiveness of a live migration scheme:  Downtime is the time interval during which services are entirely unavailable [1]. It is the time from when VM pauses on the source machine to when it resumes on the destination. Synchronization is usually performed in downtime. So the synchronization mechanism impacts on downtime.  Total migration time is the duration from when the migration starts to when the states on both machines are fully synchronized [1]. Decrease the size of transferred data, e.g. to compress the transferred data before sending it, will show a reduction in total migration time.  Performance overhead is the decrement of the service performance caused by migration. It is evaluated by the comparison of the service throughput during the migration and without migration. A highbandwidth network connection between the source and the destination will decrease

117

downtime, disruption time, and migration time to a certain extent. B. Requirements for a Whole-System Live Migration

Based on the metrics discussed in section III-A, an ideal VM migration is a whole-system migration with short downtime, minimized disruption time, endurable migration time, and negligible performance overhead. And it only transfers the run-time states without any redundancy. But this ideal whole-system live migration is hard to implement. Transferring large-volume local storage incurs a long migration time. It is difficult to maintain the consistency of the storage between the source and destination during such a long migration time while retaining a short downtime. The design of our system focuses on the following requirements:  Live migration: VM keeps running during most time of the migration process. In other words, clients can’t notice that the services on the VM are interrupted during the migration.  Minimal downtime: An ingenious synchronization method is required to minimize the size of the data transmitted in the downtime.  Consistency: The VM’s file system is consistent and identical during migration except downtime.  Minimizing performance overhead: A nonredundant synchronization method and a set of simple protocols must be designed. And the bandwidth used by the migration process should be limited to ensure the performance of the services on the migrated VM.  Finite dependency on the source machine: The source machine can be shut down after migration. That means synchronization must be completed in a finite period of time.  Transparency: Applications running on the migrated VM don’t need to be reconfigured.  Minimizing migration time: This can be achieved if a part of the state data need not be transmitted. II. RESOURCE RESERVATION METHODS

Prajakta Ugale, S.S.Rahul, Sarita Sangawar , Souvik Saha, Yousuf Syed.

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com September 2014, Volume 2 Issue 4, ISSN 2349-4476

We implement resource reservation in both source and target machine so that we can perform efficient migration. A. SOURCE MACHINE

When we create a VM, we keep some threshold value for allocating memory size and number of CPU’s. Our basic aim is to reserve adequate amount of resources in source machine so that there is no need of migration. B. TARGET MACHINE

In the target machine, a dummy VM is been created .So that when the VM is been migrated from source to target , the dummy VM is shut down so that its resources are allocated to the migrated VM. III. FRAMEWORK OF LIVE MIGRATION In this section, we propose the resource reservation based live migration framework of multiple virtual machines.

B.

Reserving Resources

It implements different resources reservation strategies for both source machine and target machine, such as CPU resource and memory resource reservation in the source machine or the whole virtual machine resource reservation in the target machine. This module is very important in the live migration of multiple virtual machines to avoid migration failures because of the insufficient resources in the target machine for the migrated virtual machine. The resource reservation in the source machine is implemented by dynamically adjusting the migrated virtual machine’s CPU cycle and the memory resource. And the resource reservation in the target machine is implemented by temporarily creating particular virtual machines to occupy a certain number of system resources for the migrated virtual machines. C.

Monitoring Resources

It is responsible for monitoring the resource status of both virtual machines and physical machines, including the resource utilization, virtual machine configuration information (workload characteristics, VCPU number, memory size, image size) which is essential to make migration decisions. It can also be used to analyze the workload stability which is useful to avoid migration thrashing. IV. DESIGN OVERVIEW

Figure1. Live Migration Framework A. Decision maker

This module decides when to migrate the VM depending upon the resource status of both source machine and Virtual Machine.It controls the real migration process. It will choose the right target machine from the candidate target machine list and trigger the migration at particular time. This module relies on the Decision-Maker module, and executes the migration strategies made by the Decision-Maker module.

118

The following stages discuss the basic live migration strategy: Stage 0: Pre-SelectionTarget machine is preselected before the migration is performed depending upon the resource status of that particular target machine. Stage 1: Reservation- CPU and memory resource needs to be reserved in the target machine so that there is no migration failure. For that reason dummy VM is being created on the target machine. The dummy VM starts running but it is not providing any service to the client.

Prajakta Ugale, S.S.Rahul, Sarita Sangawar , Souvik Saha, Yousuf Syed.

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com September 2014, Volume 2 Issue 4, ISSN 2349-4476

When the VM is migrated to target machine , then the dummy VM is shut down so that its resources can be allocated to the migrated VM. Stage 2: Copy of Pages- In this phase, all the pages and including the dirty pages are copied to the target machine. Stage 3: Suspend and redirect- VM is suspended in source machine and all the traffic gets redirected to the target machine. Stage 4: Acknowledgment- After the source machine is been acknowledged that all the dirty pages has been transferred, then the source machine discards the original VM. Stage 5: Resume- The migrated VM resumes its normal operation in target machine.

V. IMPLEMENTATION DETAILS

A basic experiment was performed for live migration. First two machines were taken and Centos 6.4 was installed on both of them. Then KVM was installed to create a virtual environment on both the machine. After that image file was uploaded on both the machine. For our convenience, Ubuntu10 was chosen as VM. It was noticed that to migrate a virtual machine it is necessary to have the image file of the operating system on both the machine otherwise it is not possible to migrate. Before performing the migration it is necessary for the source machine to establish a connection with the destination machine using the IP address .Proper bandwidth is to be selected. After that live migration can be performed and it is seen that the OS on the source machine stops functioning completely but it is working properly on the destination machine. The downtime was measured approximately to be around 15secs.For resource statistics, we are considering three factorsidle% of CPU, amount of free memory used and amount of virtual memory used. We are assuming a threshold value for each factor. Each factor is compared with the threshold value. We are running a java program in background which initially analyses the status of both source and VM. When the source machine is in critical state, we analyze which VM to migrate. After this decision is made, we analyze the status of each target machine and select the best

119

target machine. Then we migrate the VM to the selected target machine. VI. CONCLUSION AND FUTURE WORK

Live migration of virtual machines is an efficient technology used to implement energy saving and load balancing in virtualized cloud computing data centre. In this paper, we study the live migration efficiency of multiple virtual machines from experimental perspective and investigate different resource reservation methods in the live migration process as well as other complex migration strategies such as parallel migration and workload-aware migration. Experimental results show that: (1) Live migration of virtual machine brings some performance overheads. (2) The performance overheads of live migration is affected by memory size, CPU resource, and the workload types. (3) Resource reservation in target machine is necessary to avoid the migration failures. (4) The adequate system resources in the source machine can make more parallel number of migrations and can obtain better migration efficiency. (5) The workload-aware migration strategy can efficiently improve the performance of migrated workload. Based on the experimental discoveries, three optimization methods, optimization in the source machine, parallel migration of multiple virtual machines, and workload aware migration strategy, are proposed to improve the migration efficiency. Future work will include designing and implementing intelligent live migration mechanism to improve the live migration efficiency in the multiple virtual machines scenario and studying the migration strategies as an optimization problem using mathematical modeling methods. REFERENCES [1 ] Y. Luo, B. Zhang, X. Wang, Z. Wang, Y. Sun, and H. Chen, “Live and incremental whole-system migration of virtual machines using block-bitmap,” in Proceedings of the IEEE International Conference on Cluster Computing, 2008, pp 99–106. [2] H. Liu, H. Jin, X. Liao, L. Hu, and C. Yu, “Live migration of virtual machine based on full system trace and replay,” in Proceedings of the 18th ACM

Prajakta Ugale, S.S.Rahul, Sarita Sangawar , Souvik Saha, Yousuf Syed.

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com September 2014, Volume 2 Issue 4, ISSN 2349-4476

international symposium on High performance distributed computing, 2009, pp. 101–110. [3] C. Clark, K. Fraser, S. Hand, J. Hansen, E. Jul, C. Limpach, I. Pratt, and A.Warfield, “Live migration of virtual machines,” in Proceedings of the 2nd conference on Symposium on Networked Systems Design & Implementation-Volume 2, 2005, p. 286.

120

[4] M. Zhao and R. Figueiredo, “Experimental study of virtual machine migration in support of reservation of cluster resources,” in Proceedings of the 3rd international workshop on Virtualization technology in distributed computing, 2007, p. 5. [5] M. Nelson, B. Lim, and G. Hutchins, “Fast transparent migration for virtual machines,” in Proceedings of the annual conference on USENIX Annual Technical Conference, 2005, p 25.

Prajakta Ugale, S.S.Rahul, Sarita Sangawar , Souvik Saha, Yousuf Syed.