Você está na página 1de 6

2013 International Conference on Cloud & Ubiquitous Computing & Emerging Technologies

A Technical Review for Efficient Virtual Machine


Migration
Sangeeta Sharma , Meenu Chawla
Department Computer Science and Engineering
Maulana Azad National Institute of Technology
Bhopal, 462051, India
sanjsharma29@gmail.com, chawlam@manit.ac.in

target machine. Another is live migration in which user is able


to continuously execute during the migration.
Migration of entire operating system and all of its
applications as one unit (i.e. virtual machine) is less complex
than process level migration. To avoid the dependency
between the physical host machine and virtual machine,
virtual machine monitor (hypervisor) is used as an interface. It
is simply a software layer between host operating system and
virtual operating system. One of the widely used virtual
machine monitor is Xen [1]. It is a high performance resource
managed virtual machine monitor, which allows various
virtual machines to share common hardware in a safe and
resource managed fashion, but without sacrificing the
performance and functionality.
Migration of entire virtual machines means that in-memory
state can be transferred in a consistent and efficient fashion.
This includes internal states of kernel level and application
level. Live virtual machine migration is an extremely powerful
tool for the cloud managers. Two important parameters to
describe the performance of the migration are:
x Total migration time which shows the total time
needed for migration of virtual machine and to start it
on target machine.
x Downtime is duration of time for which a user waits to
resume the virtual machine.
It is required to minimize both downtime and total
migration time for the high performance of live migration.
In the migration process, complete state of running virtual
machine has to be transferred. The state of virtual machine
includes the permanent storage (i.e. the disks), volatile storage
(the memory), the state of connected devices (such as network
interface card), and the internal state of the virtual CPU.
Generally the permanent memory is provided by means of
network attached storage (NAS); so it is no required to move
it. The internal state of the VCPU and connected devices are a
few kilobytes of data and can be easily sent to the target host.
So, the main focus of live migration is efficient transfer of
volatile memory which may be up to several gigabytes.
This paper presents the different techniques to improve the
performance of live migration by transferring the memory
efficiently. It can be optimized by reducing the total migration
time and downtime but encompasses considerable overhead.
Memory transfer can be categorized into three phases [2]:

AbstractThis paper presents the recent technical research


survey on the efficient live migration of virtual machines. Virtual
machine migration is required for many reasons like load
balancing, energy reduction, dynamic resizing, and to increase
availability. Live migration is an important feature of
virtualization technology. Virtualization technology has gained
toehold with ascend of cloud computing. Simply, virtualization
runs multiple virtual machines on the single physical machine.
Virtual machine is a software based component which is an
abstraction of the underlying hardware provided by the
virtualization technology. Availability can be efficiently achieved
by managing the virtual machines properly. For this, there is a
need to migrate virtual machine from one physical machine to
another, without expending long migration time and downtime.
Virtual machine is running, while the migration is in progress is
known as live migration. There are many techniques proposed to
enhance the live migration, and reduce its duration
This paper reviews the different approaches which have been
proposed by researchers to improve the live migration by
decreasing the downtime and total migration time. This paper
can be useful for the researchers to work on this field.
Keywords live migration; cloud computing; virtualization;
xen; downtime.

I.
INTRODUCTION
Cloud Computing provides the sharing of multiple
resources between several users at the same time by the use of
internet. It is also highly scalable, elastic in nature. It works
on pay as you use model.
Virtualization technology can play an important role to
provide the multi-tenancy effectively. Virtualization gives an
abstract view of hardware by means of virtual instance of
multiple guest operating systems on a single host operating
system. This virtual instance of operating system is known as
virtual machine. The user can access the hardware in terms of
virtual machines, which can simulate the hardware. As many
virtual machines can be created on a single physical machine,
depending on their capacity. It is clear that by proper
management of virtual machines can improve the resource
utilization and availability.
Virtualization allows a migration of virtual machine from
one physical machine to another. Migration can be offline
where user is halted till the virtual machine can resume on
978-0-4799-2235-2/13 $31.00
$26.00 2013 IEEE
DOI 10.1109/CUBE.2013.14

A. Memory Transfer Phases:


20

1) Push phase: Certain pages are pushed in advance to


the new destination before the source virtual machine moves
to destination. Some modified pages are resent to ensure
consistency.
2) Stop and copy phase: It is a simple way in which
source virtual machine is stopped at source and complete
virtual machine is copied or moved to destination and then
resumed it at destination.
3) Pull phase: In this the new virtual machine starts
execution on destination machine and if it needs a page that
has not been copied, then this page is pulled across the
network from the source virtual machine.
Based on these phases, two main migration approaches are
pre-copy and post-copy approaches. Pre-copy approach is a
combination of push phase with stop and copy phase; and
post-copy approach is a combination of pull phase with stop
and copy phase.
The main objective of this literature survey is to give
proper review of existing live migration techniques. So the
researchers can contribute in the development of more
efficient live migration strategy.
Rest of the paper is organized as follows: Section II
provides the short introduction to pre-copy approach and its
variants. It also includes different techniques based on this
approach. Section III introduces the post-copy approach with
its variants and different techniques based on this. Section IV
describes other techniques on this area. Finally section V
concludes the paper.

Figure 1. Pre-copy Approach

In this way total data transferred during migration processes


can be reduced considerably. In order to detect the identical
pages between storage blocks and memory, it maintains an
updated memory map of identical pages by tracing all
input/output operations to the shared storage. It also sends the
memory to disk mapping of these pages which exists on the
shared storage to the target host, from where the data is fetched
by a background process.
Jin et al. [4] proposed an optimized pre-copy algorithm to
improve the performance of virtual machine live migration.
The performance of pre-copy mechanism is dependent upon
the virtual machines memory writing speed. Memory dirty rate
should be controlled according to the available transferring
bandwidth. This approach controls the memory writing speed
through controlling the CPU scheduler of the Virtual machine
monitor. This approach claims the reduction in downtime up to
88% as compared to basic pre-copy approach with an
acceptable overhead. It works well when network bandwidth
between migration parties is low. The method improves the
migration liveness at the cost of application performance.
Ma et al. [5] improve the basic pre-copy approach by
introducing a bit-map page. Bit-map page used to mark
frequently updated pages. These frequently updated pages are
transferred in the last round of iteration in the migration
process. It ensures that these frequently updated pages are
transferred just once in the migration processes. In this way
total transferred data can be reduced. It also shows the
improvement in the total migration time as compared to precopy approach.
Alamadari et al. [6] proposes an algorithm based on one of
the cache replacement policy which is named as reuse
distance. Based on calculated reuse distance by the algorithm,
it makes decision for transferring dirty pages during each
iterated migration phase. Number of distinctive pages between
two consecutive updating of the same pages is the reuse
distance of a virtual machine page. Figure 2 shows a sequence
of updated pages and the corresponding calculated reuse
distance.
As shown in figure 2, first A is updated after three
different pages i.e. B, C, & D so; reuse distance for A is three.
A is again updated consecutively so the reuse distance is zero.

II. PRE-COPY APPROACH


In the pre-copy approach all the memory pages are send
before the virtual machine is resume to execute on the new
host machine. The pre-copy is an iterative algorithm shown in
figure 1. In iterative push phase memory pages are sent in
advance till the complete consistent memory pages are not
moved to destination machine. Next in stop and copy phase
CPU state and remaining inconsistent pages are sent to the
destination, leading it to a fully consistent state. After, each
iteration there can be some dirty pages which might have been
modified by the virtual machine, since virtual machine is
continuous running at source side. In case where updation rate
of dirty pages is faster than the rate at which pages can be
transferred to destination, stop and copy procedure must be
executed. The time to stop and copy phase is not obvious but it
is based on the trade-off between total migration time and
downtime.
Pre-copy works well for read-intensive workloads, but due
to fast page modification rate; it is not good for write intensive
workloads.
There are some techniques which try to balance or reduce
dirty memory page size to improve the performance of basic
pre-copy approach.
Jo et al. [3] present an efficient method for per-copy live
migration of virtual machine. It assumes that all physical hosts
can be attached with shared storage. The main idea behind this
method is to not send those memory pages that are also stored
on shared storage. In this method instead of all memory pages
only those memory pages are transferred which are not directly
fetched from the shared storage.

Figure 2. Sequence of Updated Pages and Corresponding Distance [6]

21

Similarly, reuse distance of B is infinite since, it is not updated


further. Similarly, reuse distance of C and D is one and two
respectively. It shows that most updated pages have small
reuse distance. When most of the updated pages have small
reuse distance then pre-copy will not work properly.
Siripoonya et al. [7] proposed a mechanism named
thread based live check-pointing (TLC). The mechanism to
save virtual machine state to a file for a later recovery is
known as virtual machine checkpointing. This method focuses
to recover virtual machine at the time of failure. It can also
reduce checkpointing overheads as compared to the traditional
approach when virtual machine has memory intensive
workloads while performing the virtual machine
checkpointing. Virtual machine thread is periodically
interrupted to hypervisor to create checkpointing thread. It
assists the virtual machine thread to incrementally copy dirty
memory pages to hash table. Thread based live check-pointing
mechanism has been illustrated in figure 3.
Ma et al. [8] proposed a method based on the idea that
virtual machine memory is not fully used. Some of the
memory is unused or free, so it is unnecessary transfer, which
leads to a waste of network bandwidth. It presents an approach
named, Memory Exploration and Encoding (ME2). It works in
twofold; first explores useful pages and then uses RLE (Run
Length Encode) algorithm to compress for further decreasing
data amount. Through results, it shows that it efficiently
decrease the total transferred data, total migration time and
migration downtime compared with Xens pre-copy algorithm.
Svard et al. [9] proposed an approach in order to address
the page resend problem, named dynamic page transfer
recording. This algorithm orders the less frequently written
pages as high priority and frequently updated pages as low
priority. This ordering help to reduce the risk of re-transfer the
frequently updated pages by saving it till last.
To achieve this there is a need to find the most frequently
updated and less updated pages. For this it calculate a page
weight for each memory page. The weight of the page depends
upon the number of times a page has been updated during the
migration. The page can be transmitted according to the page
weight. In order to speed up the process page weight can be
handled by blocks of pages instead of single pages.
To further improve the migration processes it reduces the
amount of data transferred. In order to achieve this it combines

the idea of delta compression with the reordering algorithm.


The delta compression sends the XOR of the difference
between the page versions instead of the full page contents. In
this way this algorithm increases the migration throughput and
as well as reduce migration downtime effectively and
efficiently.
Hu et al. [10] presents a method to reduce the frequently
updated and repeatedly transmitted dirty pages. For this it uses
historical statistics to identify the frequently updated pages
and transmit them in the last round of iteration. High dirty rate
page is the page which is updated in this iteration and was
modified in previous iterations more than a given number of
times. It takes the advantage of principle of locality and
significantly reduces the total migration time effectively. It
also improves the performance of migration, including less
number of iterations and fewer pages transferred.
Zhang et al. [11] introduces data duplication into virtual
machine migration approach, called Migration with Data
Duplication (MDD). MDD framework is shown in figure 4. It
uses hash-based fingerprints technology to identify the
similarity between memory pages in order to decrease
memory data to be transferred.
For further improvement it utilizes Run Length Encoding
(RLE) to eliminate the redundant data during migration and
saves network bandwidth. Its results give considerable
improvement on Xens default pre-copy migration algorithm,
in terms of total data transferred during migration, total
migration time and migration downtime.
III.

POST-COPY APPROACH

In this approach first the virtual machine stopped at source


machine and transferred to destination host with its minimal
processor state and starts its execution. Thereafter, fetching of
memory pages begins as needed, from the source over the
network. Post-copy approach is illustrated in figure 5.
In post-copy approach, each memory page is transferred only
once, which is the main benefit over pre-copy approach. Precopy approach provides high reliability against system fault,
since it contains copy of the memory pages in both sides,
whereas in post-copy approach memory page can only be
found at one side only. Based on the manner in which pages
are fetched, different variants [13] of post-copy are there each
of which provides incremental improvements.

Figure 3. TLC Mechanism [7]


Figure 4. MDD Framework [11]

22

Figure 5. Post-Copy Approach

Post-copy via demand paging: It is the simplest and


slowest option between all other variants. When virtual
machine starts execution at the new host, it needs to access
memory pages not present in its memory, which leads the page
faults. These page faults can be serviced by requesting the
referenced page over the network. Servicing of each page fault
can slow down the virtual machine due to the network load.
Even if the page is transferred only once then also it results in
long resume time. Thus the performance of this variant is not
acceptable in terms of both total migration time and
downtime.

Figure 6. Design of System Components [13]

Hirofuchi et al. [13] proposed an advanced virtual machine


dynamic consolidation system through live migration
approach. This paper shows considerable improvement on
performance by the use of post-copy live migration approach.
In this paper with response to every changing resource usage,
virtual machine locations are reactively optimized. Virtual
machines are quickly switched between the hosts in order to
resolve the sudden overloading of server nodes. For this it
developed a prototype of consolidation system. The design of
consolidation system is illustrated in Fig.6.
The consolidation system consist of three main
components i.e. Load monitor, Relocation Planner and Virtual
Machine Controller. Load Monitor collects resources usage
statistics from each server node, at every one second and put it
into a database. Relocation Planner calculates a relocation
plan, by analyzing the resource usage histories from the
database. Virtual Machine Controller activates live migration
to server nodes according to the results from relocation
planner. Its result shows a higher degree of performance
assurance than by using pre-copy migration approach.

1) Post-copy via active pushing: In order to reduce the


page fault, another variant of post copy approach is active
pushing. In this the pages are proactively push to the
destination even as the virtual machine continues executing at
the destination. Major faults can be served concurrently as
demand paging. Active pushing avoids those page faults
which are already transferred.
2) Post-copy via pre-paging: The basis of this method is
to transfer those pages which can be accessed by the virtual
machine in the near future. It works by using the faulting
addresses to predict the spatial locality of virtual machine
memory access pattern. In order to reduce the page fault, this
will increase the probability that the pushed page can be
accessed by the virtual machine in the near future.
3) Hybrid pre and post-copy: This variant uses the
benefit of both pre-copy and post-copy approaches. In this
approach at the first iteration it works as pre-copy approach,
while virtual machine running on source machine. After first
iteration of memory transfer, the virtual machine is stopped
and resumes at destination machine with its processor state
and dirty pages. Then, remaining pages are transferred as postcopy approach. It gives better total migration time for both
read-intensive workloads as well as write-intensive workloads.
In order to optimize the live migration on the basic of postcopy approach, some of the techniques have been discussed:
Hines et al. [12] shows the improvement in total migration
time, pages transferred and network overhead by using postcopy approach with adaptive pre-paging. Post-copy approach
with adaptive pre-paging is used to eliminate all duplicate
page transmission. On the other side a dynamic selfballooning mechanism is used for further improvement.
Dynamic self-ballooning (DSB) mechanism eliminates the
transfer of free memory page. DSB periodically releases free
pages in a guest virtual machine back to the hypervisor. It
significantly speeds up migration with negligible performance
degradation.

IV.

OTHER TECHNIQUES

In order to increase the performance of virtual machine


migration, this section shows some other techniques which
work well in this area.
Luo et al. [14] presents the scheme for a whole-system
migration, which includes memory data, CPU state, and local
disk storage of the virtual machine. To optimize the wholesystem migration, it proposes a three-phase migration
algorithm based on the incremental migration approach. It
uses block-bitmap to track all the updates to the local disk
storage during the migration. Three-phase algorithm reduces
the total migration time and downtime, as well as maintains
disk storage data integrity and consistency. Three phases of
the algorithm are pre-copy, freeze-and copy, and post-copy.
The process of the algorithm is shown in figure 7.
Jin et al. [15] presents an adaptive memory compression
approach for the live migration of virtual machine named,
Memory Compression based Virtual Machine Migration
(MECOM). It designs adaptive zero-aware memory
compression algorithm for the virtual machine live migration
process, based on the memory page characteristics. It says that
compression is useless if compression time exceeds the data
transfer time of each round in migration processes. It shows
that instead of simply compressing the memory pages, an
appropriate compression algorithm based on the memory page

23

the source host. Recorded trace file is used to synchronize the


migrated virtual machines execution state, by iteratively
transferred it to the target host. Its result shows that it
significantly reduces the migration downtime and network
bandwidth consumption as compared to precopy-based
migration algorithm.
Li et al. [20] propose a live migration strategy for the
virtual machine based on the performance predicting
algorithm. For this it uses an Optimal Singular Value
Decomposition (OSVD) algorithm. This algorithm analyze the
current status of system performance information, such as
CPU, memory, disk I/O, network and predict the result of the
next moment. Based on the result, it decides whether a
migration is triggered, and to which virtual machine or where
it will be migrated. Its result shows that it achieves effective
performance prediction and also reduces the number of
unnecessary migrations. It improves the system application
performance by avoiding the system into overload state. It also
maintains the system stability and achieves the loadbalancing.
Ye et al. [21] works with different resource reservation
methods and focuses on the live migration strategy of multiple
virtual machines. With proper analysis it shows the
performance overhead that occurs due to the live migration of
virtual machines. It also gives the impact of resource
reservation method on the performance of migration process.
It also analyzes the effect and efficiency of parallel live
migration of multiple virtual machines and workload aware
live migration strategies. It presents the framework for live
migration of multiple virtual machines with resource
reservation technology. Migration Controller, Resource
Monitor, Migration Decision Maker and Resource Reservation
Controller are the main modules of the framework.
This method reserves the resources on the both sides
source as well as target side. At the source side resource
reservation in Xen is done by adjusting the CAP value to limit
the maximum utilization of the resource. Extra virtual
machines for CPU bound workloads at the target are created to
reserve the resource at the target side. The live migration
framework of multiple virtual machines with resource
reservation technology is shown in figure 8.

Figure 7. Three-Phase Whole-System Migration [14]

characteristics should be used which gives lossless


compression as well as small overhead.
Liu et al. [16] present a slowdown scheduling algorithm
for Xen live migration by which it introduces VCPU (Virtual
CPU) activity adjusting scheme. In this it is shown that by
decreasing the CPU resources, the corresponding CPU (VCPU
for migration domain) activity will be reduced. This is done
until the dirty rate is lowered down to meet the stop and copy
condition. Due to reduced access to CPU resources, the dirty
rate will be dropped significantly. In this way total migration
time and downtime for virtual machine migration is reduced,
but the impact on the performance of an application has not
been analyzed.
Park et al. [17] proposes an efficient check pointing
technique of virtual machine in terms of space and fast
execution. To achieve this it eliminates the redundant data and
saves only a subset of virtual machine memory pages. At the
hypervisor level it transparently intercepts the I/O request and
maintains an up-to-date mapping of memory pages in
page_to_block map. Those pages which are written into disk
are excluded at the checkpoint time. It shows a significant
reduction in the stored data, as well as the time required to
complete the check-point.
Svard et al. [18] shows a delta compression approach in
order to migrate the large virtual machines and heavy load
virtual machines with a shorten migration time and downtime.
It also enables live migration with a reduced risk of service
interruption. This method stores only the changes between the
versions instead of full data sets. This data can be further
compressed by binary run length encoding (RLE) in order to
increase the performance of migration process, in terms of
increased migration throughput and reduced downtime. A
delta compression live migration algorithm developed as a
modification of the KVM hypervisor.
Liu et al. [19] describes a technique namely, CR/TRmotion to provide fast and transparent virtual machine
migration based on check-pointing/recovery and trace/replay
technology. It provides fast, transparent virtual machine
migration for both LAN and WAN environment. With little
performance overhead, a trace daemon continuously logs nondeterministic events of the virtual machines and recorded at

V.

CONCLUSION

This paper describes the cloud computing and its related


issues. Some issues like load balancing, resource utilization,
availability, energy conservation need proper management of
cloud resources. For properly managing the cloud resources,
virtual machine migration comes in picture. Virtual machines
can be migrated from overloaded machine to under loaded
machine. But, it will be only helpful when the efficient live
migration techniques are used. This paper shows the various
migration methods i.e. stop and copy, pre-copy and post-copy.
All these techniques are evaluated in the basis of migration
performance parameters i.e. total migration time, migration
downtime, and total data transferred in the various papers.
There is a still need for an efficient virtual machine live
migration technique, especially when the available bandwidth
is low. There is also need for an approach to implement live
wide area migration of virtual machines including local
persistent state.

24

[13]

[14]

[15]

[16]
Figure 8. Framework of multiple virtual machine live migration [21]

With the effectiveness, security issues are also arising with the
virtual machine live migration. There is a need to incorporate
security in virtual machine live migration with the
performance issue. The secure live virtual machine migration
is also an interesting research topic for future work.

[17]

[18]

REFERENCES
[1]

P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, AlexHo,


RolfNeugebauer, I. Pratt, and A. Warfield, Xen and the art of
virtualization, In Proceedings of the nineteenth ACM symposium on
Operating Systems Principles (SOSP19), pp. 164177, ACMPress,
2003.
[2] C. Clark, K. Fraser, S. Hand, J. G. Hansen, E. Jul, C. Limpach, I. Pratt,
and A. Warfield Live migration of virtual machines, In: Proceedings
of the 2nd ACM/USENIX symposium on networked systems design and
implementation (NSDI), 2005.
[3] C. Jo, E. Gustafsson, J. Son, and B. Egger, Efficient Live Migration of
Virtual Machines Using Shared Storage, VEE13, March 1617, 2013,
Houston, Texas, USA acm, pp 41-50, 2013.
[4] H. Jin , W. Gao , S. Wu , X. Shi, X. Wu, F. Zhou, "Optimizing the Live
Migration of Virtual Machine by CPU Scheduling," 1. Network and
Computer Applications, voL34, Issue 4, pp. 1088-1096, 2011.
[5] F. Ma, F. Liu, and Z. Liu, "Virtual Machine Migration Based on
Improved Pre-copy Approach," In Proc. IEEE Int'l Conf. Software
Engineering and Service Sciences, pp.230-233, 2010.
[6] J. Farzin Alamdari, and K. Zamanifar, A Reuse Distance Based
Precopy Approach to Improve Live Migration of Virtual Machines in
2nd IEEE International Conference on Parallel, Distributed and Grid
Computing, pp 551-556, 2012.
[7] V. Siripoonya, and K. Chanchio, Thread-Based Live Checkpointing of
Virtual Machines, IEEE International Symposium on Network
Computing and Applications, pp 155-162, 2011.
[8] Y. Ma, H. Wang, J. Dong, Y. Li,and S. Cheng, ME2: Efficient Live
Migration of Virtual Machine With Memory Exploration and
Encoding, IEEE International Conference on Cluster Computing, pp
610-613, 2012.
[9] P. Svard, J. Tordsson, B. Hudzia, and E. Elmroth, High performance
live migration through dynamic page transfer reordering and
compression, Third IEEE International Conference on Coud
Computing Technology and Science, pp 542-548, 2011.
[10] B. Hu, Z. Lei, Y. Lei, D. Xu, and J. Li, A Time-Series Based Precopy
Approach for Live Migration of Virtual Machines, IEEE 17th
International Conference on Parallel and Distributed Systems, pp 947952, 2011.
[11] X. Zhang, Z. Huo, J. Ma, and D. Meng, Exploiting Data Duolication to
Accelerate Live Virtual Machine Migration, IEEE International
Conference on Cluster Computing, pp 88-96, 2010.
[12] M. R. Hines and K. Gopalan, Post-copy based live virtual machine
migration using adaptive pre-paging and dynamic self-ballooning. In

[19]

[20]

[21]

25

Proceedings of ACM SIGPLAN/SIGOPS International Conference on


Virtual Execution Environments (VEE), Washington, DC (March 2009).
T. Hirofuchi, H. Nakada, S. Itoh, and S. Sekiguchi. Reactive
consolidation of virtual machines enabled by postcopy live migration,
In Proceedings of the 5th international work-shop on Virtualization
technologies in distributed computing, VTDC 11, pages 1118, New
York, NY, USA, 2011. ACM.
Y. Luo, B. Zhang, X. Wang, Z. Wang, Y. Sun, and H. Chen, Live and
incremental whole-system migration of virtual machines using blockbitmap, in Proceedings of the IEEE International Conference on Cluster
Computing (Cluster08), pp. 99106, 2008.
H. Jin, L. Deng, S. Wu, X. Shi, and X. Pan. Live virtual machine
migration with adaptive, memory compression, In Cluster Computing
and Workshops, 2009. CLUSTER 09. IEEE Inter-national Conference
on, pages 1 10, 31 2009-sept.
Z. Liu, W. Qu, W. Liu, and K. Li, Xen Live Migration with Slowdown
Scheduling Algorithm, The 11th International Conference on Parallel
and Distributed Computing, Applications and Technologies, pp 215-221,
2010.
E. Park, B. Egger, and J. Lee, Fast and space-efficient virtual machine
checkpointing In Proceedings of the 7th ACM SIGPLAN/SIGOPS
international conference on Virtual execution environments, VEE 11,
pages 7586, New York, NY, USA, 2011. ACM.
P. Svard, B. Hudzia, J. Tordsson, and E. Elmroth. Evaluation of delta
compression techniques for efficient live migration of large virtual
machines, In Proceedings of the 7th ACM SIG-PLAN/SIGOPS
international conference on Virtual execution environments, VEE 11,
pages 111120, New York, NY, USA, 2011. ACM.
H. Liu , H. Jin, X. Liao, C. Yu, and C. Xu, Live Virtual Machine
Migration via Asynchronous Replication and State Synchronization
IEEE Transactions on Parallel and Distributed Systems, vol. 22, no. 12,
december 2011, pp 1986-1999.
Z. Li, W. Luo, X. Lu, and J. Yin, A Live Migration Strategy for Virtual
Machine Based on Performance Predicting, International Conference
on Computer Science and Service System, pp 72-76, 2012.
K. Ye, X. Jiang, D. Huang, J. Chen, and B. Wang, Live Migration of
Multiple Virtual Machines with Resource Reservation in Cloud
Computing Environments, IEEE 4th International Conference on Cloud
Computing, pp 267-274, 2011.

Você também pode gostar