Documents
Resources
Learning Center
Upload
Plans & pricing Sign in
Sign Out

50120140505011

VIEWS: 0 PAGES: 12

									  INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING &
 International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
 ISSN 0976 - 6375(Online), Volume 5, Issue 5, May (2014), pp. 82-93 © IAEME
                                 TECHNOLOGY (IJCET)

ISSN 0976 – 6367(Print)
ISSN 0976 – 6375(Online)
                                                                                 IJCET
Volume 5, Issue 5, May (2014), pp. 82-93
© IAEME: www.iaeme.com/ijcet.asp                                              ©IAEME
Journal Impact Factor (2014): 8.5328 (Calculated by GISI)
www.jifactor.com




             EMPIRICAL STUDY ON OFFLINE VS LIVE MIGRATION

                                           Anala M R
            Department of Computer Science and Engineering, R V College of Engineering
                                        Bangalore, India

                                           Shobha G
            Department of Computer Science and Engineering, R V College of Engineering
                                        Bangalore, India



 ABSTRACT

         Virtualization is a state-of-the-art technology facilitating resource optimizations by providing
 an environment conducive to execute as many VMs as possible. The proliferation of VMs on a
 physical server makes the resource management convoluted. This difficulty in managing the
 resources results in these VMs not to perform optimally and seldom demonstrate poor performance.
 Often this underperformance may result in the VM to fail and stop working. Hence, it becomes
 necessary to migrate a VM from a source to a destination. When the migration decision has been
 taken, it becomes necessary to analyze the performance of applications during migration since all the
 applications will not exhibit the same performance during migration. The Migration can be
 conducted offline or live. This paper aims at analyzing the performance of offline and live migration
 techniques with respect to total migration time, downtime and performance of an application during
 migration.

 Keywords: Offline migration; Live migration; Performance; Migration time; Downtime.

 I. INTRODUCTION

         The virtualization technology’s main motivation is to run multiple and as many VMs as
 possible to execute multiple tasks. As and when the number of VMs in a server increases, this surge
 makes it difficult to manage the resources allocated to these VMs. The difficulty in resource
 management results in underperformance of VMs. These VMs may collapse and fail to continue to
 serve. To avoid breaking up of these VMs, it is necessary to migrate a running VM from source host
 to destination host for balancing the load.

                                                   82
                                                                           0976
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
                                    Issue
ISSN 0976 - 6375(Online), Volume 5, Issu 5, May (2014), pp. 82-93 © IAEME

TYPES OF MIGRATION

        Migration can be performed on the fly (live migration) or offline. In offline (stop and copy)
migration, VM running at the source is suspended and memory image is copied to the destination.
                                                                                  involved
Figure 1 illustrates the working of stop and copy migration (offline). The steps involved in stop and
copy migration is to stop VM running at source, migrate VM’s memory image from source to
destination and finally to start VM at destination.
        Live migration is a process of moving the VM from one physical machine to another, on the
fly, keeping in mind to be as less disruptive as possible. Ideally, when live migration happens under
perfect conditions, it should be seamless i.e. the whole process should happen in an end user agnostic
manner. Live migration allows an administrator to take a virtual machine offline for maintenance or
upgrading without subjecting the system's users to downtime. The goal is for an end user to not notice
the effect of live migration.




                                         1
                                  Figure 1: Stop and Copy migration

       There are many algorithmic approaches to conduct VM memory migration, but the scope of
                              copy
this paper is to discuss pre-copy memory migration [1]. This paper analyzes the live migration
                     copy
performance of pre-copy approach and offline migration for different applications. Figure 2 shows the
             re
scenario before live migration and Figure 3, after live migration.




                                     Figure 2: Before migration

                                                                         image
    Figure 2 illustrates the process of live migration. Here, the memory image of VM running at
source host is copied to destination host in iterations. When both source and destination are
synchronized i.e. maintain consistent copies, the VM image at source is destroyed and the VM
continues to run in destination as shown in Figure 3.

                                                 83
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
ISSN 0976 - 6375(Online), Volume 5, Issue 5, May (2014), pp. 82-93 © IAEME




                                       Figure 3: After migration

II. LITERATURE SURVEY

        As virtualization gains popularity for large computing environments, management of VMs is
becoming an important problem. The efficiency of the platform as well as the performance of
applications running on the platform is critically dependent on the characteristics of the applications
and the availability of required resources. If the required resources are not available at the source host
and the VM is under resource stress situation, then the VM has to be relocated for continued services.
The resource reallocation can be achieved using replication or migration. Replication allows creating
a replica of a virtual machine on another physical machine.
        The first study on replication is conducted in [2]. This study compares both replication and
migration mechanisms. It concludes that replication is preferred over migration when the CPU usage
is high since migration process consumes computational resources. If the CPU usage is relatively low,
then the migration mechanism is used.Performance evaluation of both live and non-live migration
methods, presented in [3], demonstrates that the performance of processes running on a migrating
virtual machine severely declines in virtualized computing environment. The analysis revealed that a
host OS communication and memory writing between two hosts are the main reasons for the decline.
         Live migration is a widely used technique for resource consolidation, fault tolerance, load
balancing and power saving. The research is carried out in multiple directions to achieve impressive
performance during live migration with respect to performance of an application running in VM, it’s
total migration time and the downtime experienced during live migration. The design for migrating
OSes running services with liveness constraints using the concept of writable working set is
demonstrated in [1]. Improved pre-copy approach using bitmap page to mark frequently updated
pages to ensures that frequently updated pages are transmitted only once in the iteration process is
introduced in [4]. Post-copy migration [5] defers the transfer of a VM’s memory contents until after
its processor state has been sent to the target host.
         Research is ongoing in the area of improving performance of live migration. The various
factors like total migration time, downtime, page dirty rate etc. influences the performance of live
migration. The following text discusses the progress of the work towards this direction.
         Link speed and page dirty rate [6] are the two parameters affecting the live migration
performance. The downtime is minimized in [7] using model called memory change Probability
Density Function (PDF) of the VM. The performance evaluation on the effects of live migration of
virtual machines based on the applications running inside Xen VMs are presented in [8].
         Dynamic resource management for virtual machines using live migration techniques in cloud
environment is discussed [9]. Migration heuristics are categorized to reduce power consumption and
balancing load across physical machines. The impacts of different resource reservation methods on
the performance of live migration are investigated in [10]. To improve the resource utilization a new
live virtual machine migration strategy is proposed in [11] and using the characteristics of workloads,
hotspots are detected. The selection of migrating VM and destination host depends on multi-threshold
patterns.
         To optimize consumption of energy, an approach is proposed to determine the best candidate
of migrating VM and also to choose a destination PM. The memory-compression based VM

                                                   84
                                                                           0976
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
                                    Issue
ISSN 0976 - 6375(Online), Volume 5, Issu 5, May (2014), pp. 82-93 © IAEME

                          ]
migration approach [12] to compresses memory before migration is used to lessen migration time.
                       copy                                                                  [13].
The behavior of pre-copy live migration for memory intensive applications is analyzed in [13 The
                                               [14]
performance of live VM migration is studied [14] under different levels of resource availability. Two
             oblivious                                                                          predict
application-oblivious models were designed based on workload analysis at hypervisor level to predic
                               ].           estimates
the cost of live migration [15]. This study estimates the cost of live migration based on performance
                         aware                        [16]
and energy. Affinity-aware migration technique [16] allocates resources to virtual machines
considering dynamism in network topology and job communication patterns.

III. PERFORMANCE ANALYSIS OF OFFLINE MIGRATION

    During offline migration, the running VM instance is suspended and a snapshot of this suspended
memory image is moved from the source host to the destination host. The copied VM’s memory
image is resumed on the destination host and the memory the VM used on the source host is freed.




      Figure 4: Offline migration’s Total Migration Time(TMT) and Downtime(DT) for different
                                             application

        The drawback of offline migration is due to the fact that the services provided by that VM are
suspended for an interim duration equal to the total migration time and the total migration time
                                                               .
depends on the memory size of the VM as shown in Figure 4. This section discusses the performance
of offline migration for different applications.

IV. LIVE MIGRATION APPROACHES

        Offline migration dictates that the currently running VM be suspended and this suspended
                    ration
VM’s memory image be moved from a source host to a destination host. The copied VM’s memory
image is resumed on the destination host and the memory the VM used on the source host is freed. On
the other hand, live migration facilitates migration of the VMs from a source host to destination host
on the fly. The administrators of data centres use live migration as an essential tool for high
                                               facilitates
availability of resources. Live migration facilitates high availability, fault management, load
                     level
balancing, and low-level system maintenance. Since live migration is performed on the fly, it results
in an impressive performance with minimal service downtimes. This section discusses pre copy
approach to achieve live migration.

 A.  LIVE MIGRATION USING PRE       PRE-COPY APPROACH
                                                    pre-copy
         This section discusses the idea behind pre copy approach for live migration. During live
migration of virtual machines, the hypervisor is responsible to copy all the memory pages from the
 ource
source host to the destination host while the VM is still running on the source host. The frequently
                                           re-copied.
updated pages, known as dirty pages are re copied. The recopying of dirtied pages is continued till the
           copied                             dirtyi                             copied
rate of re-copied pages is not less than page dirtying rate. When the rate of re-copied pages is less than

                                                   85
                                                                           0976
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
                                    Issue
ISSN 0976 - 6375(Online), Volume 5, Issu 5, May (2014), pp. 82-93 © IAEME

page dirty rate, the VM in source host is stopped and the remaining dirtied pages are copied to the
destination host. The VM is resumed at the destination host. The difference in time when the VM on
  e
the source host stops and VM at the destination host resumes is called downtime. The steps involved
                                   copy
in live migration of VMs using pre-copy approach are detailed below.

Step1: Preparation
         Initially, a request is issued to migrate a VM from the source host to the destination host after
confirming the availability of resources. Once the availability of resources at destination host is
confirmed, a VM of required size is reserved. If the required resources are not found at the
                                        ntinues
destination host, the VM simply continues to run on the source host unaffected. In the first iteration,
all pages are transferred from the source host to the destination host. The pages that are transferred
initially is called as the working set. Successive iterations copy only the dirtied pages from the
working set. The VM at the source host is suspended and the network traffic is redirected to the
destination host. At the end of this stage, there is a consistent suspended copy of the VM at both the
                                          The
source host and the destination host. The copy at the source host is still considered to be the primary
and is resumed in case of failure.

Step2: Migration
        After receiving a consistent OS image from the source host, a handshaking takes place
between the destination and the source hosts, the destination being the initiator and source the
responder. The source host discards the original VM and the destination host becomes the primary
host. The logical steps that are followed during the preparation and migration are summarized in
Figure 5.
                     migration
        During pre-migration process, the destination host is examined for availability of resources.
Once the resource availability is confirmed, the required resource for VM is reserved at the
                                        pre-migration process where the destination host is examined
destination host. Preparation includes pre                                    tination
for resource availability to run the VM. In reservation stage, the resources for the new incoming VM
are reserved at the destination host. Then the source VM is stopped and dirty pages are copied to the
destination host. Migration is committed when the destination host’s VM is synchronized with VM
running at the source host. Once the destination host receives the consistent copy of the VM running
                                                                                    hos
at the source host, the VM at source host is stopped and the VM at the destination host is activated.




                          Figure 5: Time line diagram for pre
                                                          pre-copy approach


                                                   86
                                                                           0976
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
                                    Issue
ISSN 0976 - 6375(Online), Volume 5, Issu 5, May (2014), pp. 82-93 © IAEME

V. EXPERIMENTAL ANALYSIS OF OFFLINE AND LIVE MIGRATION

       This section discusses the empirical analysis on performance of applications during offline
and live migration, addresses the impacts of different parameters that affect the offline and live
                                                       virtualization platform.
migration performance using pre- copy approach on xen virtualizat

   •   DOWNTIME


                                                                       >/s

                                                                       K&&>/E




                                            D           D
                           




                                          6
                                   Figure 6: Offline vs live downtime

        The downtime in offline migration is more compared to live migration as illustrated in Figure
                                                                                                   f
6. This is due to the fact that the services provided by the VM in offline migration are suspended for
an interim duration equal to the total migration time. However in pre copy migration it is suspended
only in the final iteration. The reduction in amount of service downtime is achieved using live
migration.
                                                       with
        The downtime in offline migration increases with increase in size of VM since the downtime
is equal to total migration time. In live migration, downtime may or may not increase with increase
in VM’s size, but depends on the rate at which memory pages are dirtied. Figure 6 shows the
downtime for live and offline migration which indicates that in offline migration downtime increases
with VM size, but in live migration it may or may not.

   •   TOTAL MIGRATION TIME
       The total migration time in offline migration depends only on the size of VM but in live
       on
migration total migration time depends on the size of VM and application behaviour as shown in
Figure 7. The total migration time in live migration is at least equal to total migration time of offline
migration.




                                    7:
                             Figure 7 Offline vs live total migration time



                                                   87
                                                                           0976
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
                                    Issue
ISSN 0976 - 6375(Online), Volume 5, Issu 5, May (2014), pp. 82-93 © IAEME

   •    PERFORMANCE OF APPLICATIONS
        The performance of applications during live migration is better than offline migration as
shown in Figure 8. This is due to the fact that service downtime in live migration is lower than in
offline migration.




                           W
                                          /W     Z D       W           
                                      KDWZ^^/KE  D                   D
                                         D/W^      /               t
                                                                   t
                                                                   K&&>/E
                                          8:
                                   Figure 8 Performance of application

 VI. PERFORMANCE ANALYSIS OF LIVE MIGRATION

   •   MEMORY VS DOWNTIME
       As shown in Figure 9 the downtime is not directly proportional to the size of VM, but it also
                                                                             rate
depends on the application behaviour. That is, the downtime depends on the rate at which memory
pages are dirtied, apart from the size.

                                                                   D
                                                                   D
                                                                  D
                               d




                                          K             ' ^Y>


                                            9:
                                     Figure 9 Memory vs downtime

   •   CPU VS DOWNTIME
       Figure 10 illustrates that the increase in CPU will not change the downtime; it can be
concluded that the change in amount of computational resources does not have any impact on the
downtime.




                    Figure 10: TMT and DT of VM vs Computational resource




                                                   88
                                                                           0976
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
                                    Issue
ISSN 0976 - 6375(Online), Volume 5, Issu 5, May (2014), pp. 82-93 © IAEME

   •   CPU VS TOTAL MIGRATION TIME

                                                                                 VCPU=1,RAM=1GB

                                                                                 VCPU=2,RAM=1GB

                                                                                 VCPU=1,RAM=2GB




                          Migration time-sec
                                                                                 VCPU=2,RAM=2GB

                                                                                 VCPU=1,RAM=512MB

                                                                                 VCPU=2,RAM=512MB




                                                         Cap Value in %



                                                      11:
                                               Figure 11 Total migration time vs resources

        Computational resource has no impact on total migration time and downtime assuming that
there are sufficient computational resources available to initiate live migration. Figure 11
                                                                      computational
demonstrates that the total migration time does not depend on the computational resources and
instead, depends on the size of the memory. As we can see even when computational resources are
added by keeping memory resources constant the total migration time remains nearly constant.

   •    MEMORY VS TOTAL MIGRATION TIME
                             as
        Memory resource has an impact on total migration time. As shown in Figure 12, as and when
there is an increase in size of VM, the total migration time also increases. Increase in cap value will
not alter the total migration time.




                  Figure 12: Memory vs total migrtaion with varying CPU resources

   • APPLICATION’S PERFORMANCE WITH AND WITHOUT MIGRATION
  There is a performance degradation of an application when it is migrated. The amount of
degradation differs with different application. As shown in Figure 13 the Openssl application shows
lower performance degradation compared to any other application.




                                                                   89
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
ISSN 0976 - 6375(Online), Volume 5, Issue 5, May (2014), pp. 82-93 © IAEME



                                                                   t              t




                                     KWE^^>   W    ^Y> >      dW W         /W d    /K
                                               d        W                        d   D
                      W




                  Figure 13: Performance of applications with and without migration

  • TOTAL MIGRATION TIME—BASE LINE CASE WITH APPLICATION
        The previous attempts conclude that the total migration time is directly proportional to the
size of the VM. These analyses have some loopholes. For example, applications like Openssl and
RAMspeed each occupy VM of size 512 MB but the total migration time is 49 and 206 respectively.
From this it can be concluded that the total migration time not only depends on the size of VM but
also on application behaviour.

                                                                       dDd
                                                                        dDd
                            D




                             Figure 14: TMT in offline and live migration

      Figure 14 shows the comparison of total migration time of live and offline migration. The
minimum time the live migration takes to migrate a VM is equivalent to the time taken to migrate the
same VM using offline migration.

  • TOTAL TRANSFERRED DATA VS APPLICATIONS
        The amount of data transferred during migration may not be directly proportional to the size
of the VM but it also depends on the nature of the application. The size of VM serving applications 1
and 3 is 512 MB while size of VM serving application 4 is 2 GB. From Figure 15 it can be concluded
that the total transferred data for application 1and 3 is more than the application 4. This indicates that
the size of the VM is not the only factor to decide the total transferred data, but also depends on the
nature of the application.




                                                   90
                                                                           0976
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
                                    Issue
ISSN 0976 - 6375(Online), Volume 5, Issu 5, May (2014), pp. 82-93 © IAEME




                        Figure 15: Migrated data in offline and live migration

VII. DISCUSSION AND CONCLUSIONS

        This section discusses about the findings drawn from experimental analysis. The previous
attempts show that the memory size influences downtime and migration time. They also discuss that
total migration time and downtime increase with increase in VM size. The previous attempts
fail(have not been able) to compare the amount of data transferred during migrating a VM.

The analysis of previous work conclude the following

 o Memory resource has an impact on total migration time and total migration time increases with
   increase in memory.
 o Memory resource has an impact on downtime and downtime increases with increase in
   memory.
 o Total migration time depends on link bandwidth.

The current work has been able to derive the following:

 o Computational resource has no impact on total migration time.
 o Computational resource has no impact on downtime.
 o Nature of the application influences downtime, total migration time and amount of data
   transferred during live migration.
 o The total migration time in offline migration is analysed and this would be the minimum total
   migration time for live migration.
 o The performance degradation of an application during migration also depends on the nature of
   the application. All applications do not exhibit the same level of degradation.
 o Downtime depends on memory size and application behaviour

        This paper discussed the performance analysis of offline and live migration techniques. The
offline and live migration techniques using precopy is analyzed for various parameters like total
migration time, downtime, application’s performance during migration and amount of data
transferred. It illustrates that the performance degradation during migration is dependent on type of
the application. In offline migration amount of data transferred during migration depends on the size
of the VM but in live migration the amount of data transferred during migration depends on the
nature of the application.




                                                 91
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
ISSN 0976 - 6375(Online), Volume 5, Issue 5, May (2014), pp. 82-93 © IAEME

REFERENCES

 [1]    C. Clark, K. Fraser, S. Hand, J. G. Hansen, E. Jul, C. Limpach, I. Pratt, and A. Warfield,
        “Live migration of virtual machines”, 2nd conference on Symposium on Networked Systems
        Design & Implementation - NSDI’05, Volume 2, USENIX Association, CA, USA, 2005,
        Pages 273–286.
 [2]    Gast´on Keller and HananLutfiyya, “Replication and Migration as Resource Management
        Mechanisms for Virtualized Environments”, 6th International Conference on Autonomic and
        Autonomous Systems, DOI: 10.1109/ICAS.2010.27, 2010, Pages:137-143.
 [3]    Yosuke Kuno, Kenichi Nii, Saneyasu Yamaguchi, “A Study on Performance of Processes in
        Migrating Virtual Machines”, 10th International Symposium on Autonomous Decentralized
        Systems, ISBN: 978-1-61284-213-4, DOI: 10.1109/ISADS.2011.79, 2011, Pages: 567-572.
 [4]    Fei Ma, Feng Liu, Zhen Liu, Live Virtual Machine Migration based on Improved Pre-copy
        Approach, IEEE International Conference on Software Engineering and Service Sciences
        (ICSESS), 2010, Pages: 230 – 233.
 [5]    Michael R. Hines, Umesh Deshpande, and Kartik Gopalan, “Post-copy live migration of
        virtual machines”ACM SIGOPS Operating Systems, Volume 43 Issue 3, July 2009,
        Pages 14-26 ACM New York, NY, USA, doi:10.1145/1618525.1618528.
 [6]    Sherif Akoush, Ripduman Sohan, Andrew Rice, Andrew W. Moore, Andy Hopper,
        “Predicting the Performance of Virtual Machine Migration”, IEEE InternationalSymposium
        on Modeling, Analysis & Simulation of Computer and Telecommunication Systems
        (MASCOTS), 2010, Pages:37 – 46.
 [7]    Fereydoun Farrahi Moghaddam, Mohamed Cheriet, Decreasing Live Virtual Migration
        Down-Time Using a Memory Page Selection Based on Memory Change PDF, International
        Conference      on      Networking,      Sensing    and      Control   (ICNSC),     2010,
        DOI: 10.1109/ICNSC.2010.5461517. Pages: 355 – 359.
 [8]    William Voorsluys, James Broberg, Srikumar Venugopal, and Rajkumar Buyya, Cost of
        Virtual Machine Live Migration in Clouds: A Performance Evaluation, in CloudCom, 2009,
        pp. 254-265.
 [9]    Mayank Mishra, Anwesha Das, Purushottam Kulkarni, Anirudha Sahoo, Dynamic resource
        management using virtual machine migrations Communications MagazineIEEE,
        Volume: 50 , Issue: 9 Page(s): 34- 40 .
 [10]   Kejiang Ye, Xiaohong Jiang, Dawei Huang, Jianhai Chen, Bei Wang, Live Migration of
        Multiple Virtual Machines with Resource Reservation in Cloud Computing Environments,
        IEEE International Conference on Cloud Computing, 2011 Page(s): 267 - 274 .
 [11]   Chuan, Chen , A new live virtual machine migration strategy, 2012 Information Technology
        in Medicine and Education (ITME), 2012 International Symposium on , Page(s): 173 - 176,
        Volume: 1, Aug. 2012.
 [12]   Hai Jin, Li Deng, Song Wu, Xuanhua Shi, Xiaodong Pan,Live Virtual Machine Migration
        with Adaptive Memory Compression, Cluster Computing and Workshops, 2009. CLUSTER
        '09. IEEE International Conference on, 2009-Sept. Page(s): 1 – 10.
 [13]   Khaled Z. Ibrahim, Steven Hofmeyr, Costin Iancu, Eric Roman, Optimized Pre-Copy Live
        Migration for Memory Intensive Applications, SC '11 Proceedings of 2011 International
        Conference for High Performance Computing, Networking, Storage and Analysis, Article
        No.     40 ,   ACM New       York,      NY,    USA ©2011 ,      ISBN:    978-1-4503-07710
        doi>10.1145/2063384.2063437
 [14]   Yangyang Wu Ming Zhao, Performance Modeling of Virtual Machine Live Migration, IEEE
        International Conference on Cloud Computing 2011, USA.


                                                92
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
ISSN 0976 - 6375(Online), Volume 5, Issue 5, May (2014), pp. 82-93 © IAEME

 [15] Haikun Liu, Cheng-Zhong Xu, Hai Jin, Jiayu Gong, Xiaofei Liao, Performance and Energy
      Modeling for Live Migration of Virtual Machines, HPDC '11 Proceedings of the 20th
      international symposium on High performance , distributed computing, ISBN: 978-1-4503-
      0552-5, Pages 171-182, ACM , NY, USA.
 [16] Jason Sonnek, James Greensky, Robert Reutiman and Abhishek Chandra, Starling:
      Minimizing Communication Overhead in Virtualized Computing Platforms Using
      Decentralized Affinity-Aware Migration, 39th International Conference on Parallel
      Processing Pages IEEE Computer Society Washington, 2010, DC, USA.
 [17] Dr. Narayan A. Joshi, “Load Balancing in Cloud using Process Migration”, International
      Journal of Advanced Research in Engineering & Technology (IJARET), Volume 5, Issue 4,
      2014, pp. 230 - 238, ISSN Print: 0976-6480, ISSN Online: 0976-6499.




                                            93

								
To top