6

Click here to load reader

Virtual Machine Live Migration for Pervasive Services … Machine Live Migration for Pervasive Services in Cloud ... The application of mobile cloud computing ... namely, vehicular

  • Upload
    doduong

  • View
    217

  • Download
    5

Embed Size (px)

Citation preview

Page 1: Virtual Machine Live Migration for Pervasive Services … Machine Live Migration for Pervasive Services in Cloud ... The application of mobile cloud computing ... namely, vehicular

Virtual Machine Live Migration for PervasiveServices in Cloud-Assisted Vehicular Networks

Rong Yu1, Yan Zhang2, Huimin Wu1, Periklis Chatzimisios3, and Shengli Xie11 Guangdong University of Technology, China, email: {yurong, huiminwu, eeshlxie}@ieee.org

2 Simula Research Laboratory, Norway, email: [email protected] Alexander Technological Educational Institute of Thessaloniki, Greece, email: [email protected]

Abstract—The physical resources of vehicles and roadsideinfrastructures are stringently constrained in vehicular networks.The application of mobile cloud computing technology will sig-nificantly improve the utilization of intensive physical resources.In the newly emerged paradigm of cloud-assisted vehicularnetworks, vehicle mobility poses a significant challenge to thecontinuity of cloud services. This paper proposes efficient VirtualMachine (VM) live migration mechanisms to deal with theproblem. In particular, a selective dirty page transfer strategyis designed to enhance the efficiency of data transfer in VMlive migration. Besides, an optimal resource reservation schemeis proposed to ensure sufficient physical resources at a targetcloud site such that migration dropping is significantly reduced.Simulations are carried out to demonstrate the efficiency of thetwo proposed mechanisms.

Index Terms—Mobile cloud computing, vehicular network, vir-tual machine migration, dirty page transfer, resource reservation.

I. INTRODUCTION

In the era of Internet of Things (IoT) [1], vehicular networksevolve from traditional vehicular ad hoc networks (VANETs)[2] towards the new paradigm of Internet of Vehicles (IoV)[3]. In IoV, smart sensors and actuators are equipped inthe vehicular and roadside infrastructures for data collectionand decision execution. Advanced wireless communicationtechnologies are applied for inter-networking and informationexchange. The short-range wireless communication technol-ogy, such as the Dedicated Short Range Communications(DSRC) technology, is specifically designed for the Vehicle-to-Vehicle (V2V) and Vehicle-to-Roadside (V2R) communi-cations. The IEEE 802.11p, or called Wireless Access inVehicular Environments (WAVE) [4], is currently the mostpopular international standard for DSRC. The long-rangewireless communication technology, such as 3G or 4G cellulartechnology, is exploited for remote connections between thevehicular/roadside units and the data center. Cognitive radio,as a revolutionary communication technology [5][6], could beleveraged to enrich spectrum access strategies and enhanceradio resource utilization in VANETs [7][8]. By efficientlyintegrating these heterogenous wireless communication tech-nologies, IoV constitutes a fundamental information platform,which is an indispensable part of the Intelligent TransportSystems (ITS).

There exists practical difficulty that hinders the developmentof vehicular networks. On the one hand, the volume ofdata transmitted by the vehicular and roadside sensors and

actuators are explosively growing and will reach a magnitudeof 10 ∼ 100 Tbs per day in the next few years. Thistremendous amount of data includes a variety of structuredand unstructured data, some of which should be processedrapidly for fast decision making. On the other hand, everysingle vehicle or roadside infrastructure in IoV is heavilyconstrained in the physical resource. Due to the practicalrequirements on the size, weight and cost of the hardwaresystems, the vehicles and roadside units generally have verylimited computation and storage abilities. To overcome theabove difficulty, we consider cloud-assisted vehicular networkarchitecture [9], in which cloud computing is exploited as akey enabling technology to organize all the physical resourcesin vehicular networks. However, deploying clouds in vehicularnetworks still has to face a number of challenges.

∙ Continuity of Cloud Services: Vehicles mobility leads tonetwork topology dynamic changes. Frequently changingmay interrupt ongoing cloud services. Accordingly, cloudresource management should be carefully designed.

∙ Quality of Cloud Services: Due to complicated vehicularcommunication environment, wireless channels may suf-fer from signal fading. The unstable channel quality mayseverely degrade the quality of cloud service.

∙ Accessibility of Cloud Services: Vehicle networks areessentially hybrid networks that integrate different typesof wireless mobile communication systems. The cloudarchitecture is preferred to provide interfaces to supportdifferent types of access modes.

∙ Security and Privacy of Cloud Servies: In cloud-assistedvehicular networks, important and private information aretransmitted, stored and processed in vehicles and roadsideinfrastructures which are vulnerable for underlying at-tacks. Efficient secure mechanisms are needed to protectcloud services in vehicular networks.

In this paper, we consider the development of clouds in ve-hicles, roadside infrastructures and ITS central server clusters,namely, vehicular cloud, roadside cloud and central cloud,respectively. In this paradigm, vehicles are allowed to access tothese three types of cloud sites for services. Due to the vehiclemobility, the cloud services have to shift from one cloud toanother to maintain the on-going services. Our motivation isto design efficient mechanisms to ensure the continuity ofcloud services. We propose that Virtual Machine (VM) livemigration is a key technique towards pervasive services in

2013 8th International Conference on Communications and Networking in China (CHINACOM)

978-1-4799-1406-7 © 2013 IEEE540

Page 2: Virtual Machine Live Migration for Pervasive Services … Machine Live Migration for Pervasive Services in Cloud ... The application of mobile cloud computing ... namely, vehicular

RSU-1

ITS Central Cloud

Internet

Vehicular Cloud

1

2

3

1 Inter-Roadside-Cloud Migration 2 Across Roadside-Central Cloud Migration 3 Across Central-Vehicle Cloud Migration

A A A

B

VM-AVM-A

VM-A

RSU-2

Roadside Cloud

VM-A

Fig. 1. VM live migration scenarios

vehicular networks. The remainder of the paper is organizedas follows. Section II describes the main scenarios and generalprocedure of VM live migration in vehicular networks. SectionIII proposes a selective dirty page transfer strategy to improvethe migration efficiency. A resource reservation approach isproposed and analyzed in section IV. The paper is concludedin Section V.

II. VM LIVE MIGRATION IN CLOUD-ASSISTED

VEHICULAR NETWORKS

A Virtual Machine (VM) is a simulation of a real or abstractmachine. In the literature, VM research has been mainly stud-ied in computer networks [12]. In the recent study [13], VMmigration is considered for dynamic resource management incloud environments. In [14], VM replication and schedulingare intelligently combined for VM migration across widearea network environments. The study in [15] conducts anumber of interesting experiments to compare several resourcemanagement schemes for VM migration. However, there arefew studies on VM resource management in mobile cloudenvironments. In [16], cloudlet is discussed and customizedin the mobile computing environments.

A. VM Live Migration Scenarios

VM live migration refers to the process that an operatingVM is transferred along with its applications across differentphysical machines. As shown in Fig. 1, in cloud-assistedvehicular networks, vehicles have cloud services from roadsideinfrastructures, ITS center servers and other vehicles. As avehicle moves on the road, it has to handoff to differentinfrastructures. At the same time, the cloud service has to shiftfrom one cloud site to another. Since VM is the fundamentalentity that carries out cloud services, VM live migration isa preferred means to guarantee service continuity. In a VMmigration, VM image has to be copied from the source to thedestination cloud site. In cloud-assisted vehicular networks,VM live migration has several scenarios as explained below.

∙ Inter-Roadside-Cloud Migration: In Fig. 1 (see case-1),when vehicle 𝐴 moves from the radio range of RSU-1 tothat of RSU-2, a VM migration is needed. Guest VM-𝐴

should be transferred between the two roadside cloudletsto resume its service.

∙ Across Roadside-Central Cloud Migration: In Fig. 1 (seecase-2), when vehicle 𝐴 moves out of the radio rangeof RSU-2. No more roadside cloud but central cloud isavailable. In this case, guest VM-𝐴 has to be migratedfrom roadside cloud to central cloud. After that, vehicle𝐴 will resume its service by accessing to the central cloudusing cellular wireless communications.

∙ Across Central-vehicular Cloud Migration: In Fig. 1(see case-3), when vehicle 𝐴 moves to the radio rangeof other vehicles, say, vehicle 𝐵 here, vehicle 𝐴 hasthe opportunity to transfer its VM to vehicle 𝐵. Aftermigration, vehicle 𝐴 will resume its service by accessingto the vehicular cloud using V2V communications.

∙ Across Roadside-Vehicular Cloud Migration: This sce-nario, which is not shown in Fig. 1, is similar to the oneacross roadside-central cloud, except that the destinationvehicle should potentially connect to the source RSU sothat there exists a data link for VM migration.

B. VM Live Migration Flow Chart

A VM live migration involves a number of interactionsamong the end-user vehicle, the source and the target cloudsites. We consider the pre-copy approach [10] and design theprocedure of VM live migration in cloud-assisted vehicularnetworks, as illustrated in Fig. 2.

1) RSS Threshold Detection: The end-user vehicle willperiodically monitor the Received Signal Strength (RSS)from both the source and the target cloud sites. As longas the difference of the RSS from the two cloud sitesreaches the preset threshold, the end-user vehicle willsend a migration request to the target cloud site.

2) VM & Network Resource Reallocation: Upon receivingthe VM migration request, the target cloud site shouldperform VM and network resource reallocation to decidewhether the migration is acceptable. If the migration isadmitted, a wireless link and a certain amount of cloudresource will be assigned for the migrated VM. If the

541

Page 3: Virtual Machine Live Migration for Pervasive Services … Machine Live Migration for Pervasive Services in Cloud ... The application of mobile cloud computing ... namely, vehicular

Source Cloud Site End-User Vehicle Target Cloud Site

RSS Threshold Detection

VM Migration Instruction

VM Memory Transfer

VM Migration Request

VM & Network Resource

Re-allocationVM Migration Decision

VM Service Halt

VM Memory Dirty Pages

VM Migration Completion

Radio Link Setup Request

Radio Link Setup Response

VM Services

t1

t2

t3

t4

t5

t7

t8

t9

VM Service Resume

t6Last Dirty Pages

Fig. 2. VM Live Migration Procedure

migration is denied, the end-user vehicle has to seek forother target cloud sites.

3) VM Data Transfer: After the end-user triggering themigration, the source cloud site starts to transfer VMdata to the target cloud site. The process has two steps,firstly to transfer VM memory and then to transfer VMmemory dirty pages. VM memory is the main body ofthe VM. During VM memory transfer, VM memorydirty pages are generated as the VM service is stillrunning. Generally, it will take a number of phases totransfer the dirty pages, because new dirty pages maybe generated while transferring the current dirty pages.

4) VM Service Halt: If the last dirty pages are smallenough, the source cloud site will temporally suspendthe VM service, finish the transfer of the last dirty pages,and announce the completion of VM live migration.

5) VM Service Resume: Upon VM migration completion,the end-user vehicle will request the target cloud site tosetup a new radio link, through which the VM serviceis called to continue.

In the above procedure, the total migration time starts from𝑡4 to 𝑡9, and the VM downtime starts from 𝑡6 to 𝑡9 (see Fig. 2).The duration of 𝑡5 to 𝑡6 is consumed for dirty page transfer.

III. SELECTIVE DIRTY PAGE TRANSFER FOR VM LIVE

MIGRATION

A. Selective Dirty Page Transfer Strategy

In the entire process of VM data transfer, the transfer ofdirty pages is essential to both the total migration time and theVM downtime. In the most popular VM live migration strategythat is used in Xen [11], dirty pages are completely copied tothe target cloud site in each round of transfer. This action isrepeated until one of the following conditions is satisfied:

∙ The number of dirtied pages is less than 50 in the lastround of transfer.

∙ There are 30 rounds of transfer carried out.∙ The total amount of transferred data is 3 times of the

memory size of the migrated VM.

The strategy is inefficient in the sense that most of the pre-viously transferred pages have to be retransferred later. Someof them may be transferred many times. Existing researcheshave indicated that pages in a memory are used in differentfrequency, which implies that pages should be transferred in aspecifically designed sequence. Motivated by this observation,we propose the selective dirty page transfer strategy, whichhas the main features as described below.

∙ In each round of dirty page transfer, only a fixed numberof pages are copied to the target site.

542

Page 4: Virtual Machine Live Migration for Pervasive Services … Machine Live Migration for Pervasive Services in Cloud ... The application of mobile cloud computing ... namely, vehicular

∙ The pages to be transferred are selected from all dirtypages by evaluating their dirtied rates.

∙ The dirtied rates of pages are dynamically updated ac-cording to the actual observations.

The details of the selective dirty page transfer strategy areexplained in the upcoming subsection, where we formulatethe problem into a optimal stopping framework and discussthe efficient solution.

B. Optimal Stopping Formulation and Solution

Consider a VM with 𝐼 memory pages, and let 𝑝𝑖 (𝑖 =1, 2, ⋅ ⋅ ⋅ , 𝐼) denote the 𝑖-th page in the memory. There aretwo types of pages: clean pages and dirty pages. Dirty pagesare further classified into selected dirty pages for transferand remaining dirty pages. Let Ω𝑐,Ω𝑠,Ω𝑑 denote the sets ofclean pages, selected dirty pages, and remaining dirty pages,respectively. The entire process of dirty page transfer is dividedinto a series of stages, which is indexed by 𝑛 = 1, 2, ⋅ ⋅ ⋅ . Ineach stage, which takes a time duration 𝑡𝑠, a fixed number of𝐾 dirty pages are transferred. Let 𝑥𝑖

𝑛 denote the state of the 𝑖-th page at the 𝑛-th stage (i.e., at the beginning of the 𝑛-th roundof data transfer). Here, 𝑥𝑖

𝑛 = {0, 1}, and 0 stands for clean and1 for dirty, respectively. The state vector of the VM memoryat the 𝑛-th stage is represented by 𝑋𝑛 = (𝑥1

𝑛, 𝑥2𝑛, ⋅ ⋅ ⋅ , 𝑥𝐼

𝑛).The state space is denoted by 𝒳 with size 2𝐼 .

If the dirty page transfer iteration stops at a certain stage,the VM is suspended and then the remaining dirty pages arecopied to the target site. Let 𝑌 𝑑𝑡

𝑛 and 𝑌 𝑚𝑡𝑛 , respectively, denote

the VM downtime and total migration time if the dirty pagestransfer stops right before the 𝑛-th stage. Let 𝑤 denote thedata transfer rate in terms of pages per second. We have

𝑌 𝑑𝑡𝑛 (𝑋1, 𝑋2, ⋅ ⋅ ⋅ , 𝑋𝑛) =

∣Ω𝑑,𝑛∣+ ∣Ω𝑠,𝑛∣𝑤

𝐾 (1)

𝑌 𝑚𝑡𝑛 (𝑋1, 𝑋2, ⋅ ⋅ ⋅ , 𝑋𝑛) = 𝑛𝑡𝑠 + 𝑌 𝑑𝑡

𝑛 (2)

where ∣Ω∣ represents the number of elements in set Ω. The VMdowntime and total migration time are two essential metricsfor VM live migration. Different VM services have differ-ent requirements on them. For instance, in storage-intensiveapplications, the total migration time should be reduced toprevent from additional memory occupation in both source anddestination sites; while in real-time applications, the downtimeshould be strictly limited for satisfying Quality of Service(QoS) provisioning. To balance downtime and migration timecosts, we define the cost function for stopping dirty pagetransfer right before the 𝑛-th stage by

𝑌𝑛(𝑋1, 𝑋2, ⋅ ⋅ ⋅ , 𝑋𝑛) = 𝛼𝑌 𝑑𝑡𝑛 + (1− 𝛼)𝑌 𝑚𝑡

𝑛 (3)

where 0 < 𝛼 < 1 is a constant preset by the migrated VM.If the selected dirty pages are transferred at a certain stage,

some of the clean pages will become dirty. The dirtied rate ofa given page is defined as the probability that the clean pagewill become dirty in one stage. The dirtied rate of the 𝑖-thpage is denoted by 𝑟𝑖. To select out the pages for transfer, wesort the dirty pages by their dirtied rates. At the 𝑛-th stage,the 𝐾 pages with the smallest dirtied rate are included into

the set of Ω𝑠,𝑛. Let Ω𝑛𝑒𝑤𝑑,𝑛 denote the set of the newly dirtied

pages in stage 𝑛. We have

E{∣Ω𝑛𝑒𝑤𝑑,𝑛 ∣} =

𝑝𝑖∈Ω𝑠,𝑛∪Ω𝑐,𝑛

𝑟𝑖 (4)

and then,

E{𝑌 𝑑𝑡𝑛+1(𝑋1, 𝑋2, ⋅ ⋅ ⋅ , 𝑋𝑛)} =

∣Ω𝑑,𝑛∣+E{∣Ω𝑛𝑒𝑤𝑑,𝑛 ∣}

𝑤𝐾 (5)

E{𝑌 𝑚𝑡𝑛+1(𝑋1, 𝑋2, ⋅ ⋅ ⋅ , 𝑋𝑛)} = (𝑛+ 1)𝑡𝑠 +E{𝑌 𝑑𝑡

𝑛+1} (6)

Substituting (5) and (6) into (3) yields

E{𝑌𝑛+1(𝑋1, 𝑋2, ⋅ ⋅ ⋅ , 𝑋𝑛)}

= (𝑛+ 1)(1− 𝛼)𝑡𝑠 +∣Ω𝑑,𝑛∣+

∑𝑝𝑖∈Ω𝑠,𝑛∪Ω𝑐,𝑛

𝑟𝑖

𝑤𝐾

(7)

The optimal cost function 𝑉 (𝑛) is derived by

𝑉 (𝑛) = min{𝑌𝑛(𝑋1, ⋅ ⋅ ⋅ , 𝑋𝑛),E{𝑉𝑛+1(𝑋1, ⋅ ⋅ ⋅ , 𝑋𝑛)}}(8)

where

E{𝑉𝑛+1(𝑋1, 𝑋2, ⋅ ⋅ ⋅ , 𝑋𝑛)}=

𝑋∈𝒳𝑉𝑛+1(𝑋1, 𝑋2, ⋅ ⋅ ⋅ , 𝑋𝑛, 𝑋𝑛+1 = 𝑋) (9)

For a practical VM live migration, the total migration timeis generally restricted in a predefined value. There exists amaximum number of stages before which the iteration ofdirty page transfer should be stopped. As a consequence,the optimal stopping problem we discuss here is a finitehorizon problem, which, in principle, could be solved by usingbackward recursion approach. However, the size of the statespace in our problem is large (e.g., for a VM memory of 256MB and 4 KB per page, there are 65536 pages totally) andthe state transition matrix is too huge to efficiently carry outbackward recursion algorithm. Instead, we adopt the one-step-look-ahead approach to approximate the optimal solution ofthe formulated optimal stopping problem. It turns out that thesub-optimal dirty page transfer strategy is to compare the costfunction 𝑌𝑛 with the expected cost 𝐸{𝑌𝑛+1} at each stage,and stop the transfer iteration if 𝐸{𝑌𝑛+1} > 𝑌𝑛.

To evaluate the proposed selective dirty page transfer strat-egy, we carry out a simulation to compare the performance ofthe strategy in Xen and that of our strategy. In the simulation,the migrated VM has a memory of 256 MB with 4 KB perpage and totally 65536 pages. The transfer rate is 10 MBps. Asshown in Fig. 3, the selective dirty page transfer strategy hasa prominent advantage over the Xen-based transfer strategy.Specifically, in the first 10 seconds, the amount of transferreddirty pages in the selective transfer strategy (14 K) is 3.2 timesof that in the Xen-based strategy (4 K).

IV. OPTIMAL RESOURCE RESERVATION FOR

VM LIVE MIGRATION

A. Resource Reservation Scheme

The proposed VM migration procedure (see Fig. 2) involvesresource re-allocation in the target cloud site. If the resourcesof the target cloud site have been intensively occupied, after

543

Page 5: Virtual Machine Live Migration for Pervasive Services … Machine Live Migration for Pervasive Services in Cloud ... The application of mobile cloud computing ... namely, vehicular

0 10 20 30 40 504.8

5

5.2

5.4

5.6

5.8

6

6.2

6.4

6.6x 10

4

Times (s)

Num

ber

of D

irty

Pag

es

Selective TransferXen−based Transfer

Fig. 3. The amount of remaining dirty pages in VM memory

VM migration and resource re-allocation, some of the VMsmay not have sufficient resources and may not even resumetheir services. In order to avoid resource over-commitment,the target cloud site has to deny the VM migration so asto maintain the services of existing VMs. In this case, thecloud service of a vehicle with VM migration is said to bedropped. To reduce service dropping, we propose a resourcereservation scheme. In the scheme, a small portion of thecloud site resources are reserved merely for migrated VMs,but not for local VMs. Since there are dedicated resourcesfor VM migration, the dropping rate of cloud services will besignificantly decreased.

In the proposed resource reservation scheme, resources aredivided into two categories: reserved resources and commonresources. Let 𝐶𝑟 and 𝑀𝑟 denote the reserved resources, and𝐶𝑐 = 𝐶 − 𝐶𝑟 and 𝑀𝑐 = 𝑀 − 𝑀𝑟 the common resourcesin computation and storage, respectively. In a VM migration,VM arrival refers to the event that VM is created either for anew local VM or a migrated VM. VM departure refers to therequest of VM deletion, either for an ending of VM service ora VM migration to another cloud site. The resource reservationscheme has operations in the following cases.

∙ Local VM arrival: When there is a request for creatinga new local VM, resource allocation will be carriedout, e.g., using the proposed game-theoretic allocationscheme. Since a part of resources are reserved, the localVMs can only share the common resources, i.e., 𝐶𝑐 and𝑀𝑐 in computation and storage resources, respectively. Ifthe resource allocation result satisfies all existing VMs,the new local VM is admitted; otherwise, it is blocked.

∙ Local VM departure: The resource allocation is alsoperformed when the service of one of the local VMs endsor migrates to another cloud site.

∙ Migrated VM arrival: Upon a request of VM migration,the target cloud site will re-allocate resources. In thiscase, the reserved resources will be also taken intoaccount. Specifically, the existing local VMs and the mi-grated VM will share the total amount of resources. After

0.1 0.15 0.2 0.25 0.30

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1x 10

−3

Arrival Rate of Local VMs

Rd

Reservation (Rbc=0.01)

Reservation (Rbc=0.02)

No reservation

Fig. 4. Dropping rate in terms of local VM arrival rate

re-allocation, if all the VMs (including the migrated VM)are satisfied, the VM migration is approved; otherwise,VM migration request is rejected.

∙ Migrated VM departure: The resource allocation alsohappens when the service of a migrated VM ends ormigrates to another cloud site. It is noticeable that, if thereis no migrated VMs in a cloud site, the resource allocationcould only use common resources. The reserved resourceswill be conserved for further usage upon VM migration.

B. Optimal Resource Reservation

We consider 𝐾 classes of VMs. Let 𝑐𝑘 and 𝑚𝑘 repre-sent the amount of required resources of the 𝑘-th class ofVMs in computation and storage, respectively. Let 𝑛𝑙

𝑘 and𝑛𝑔𝑘 denote the number of local and migrated VMs of class

𝑘, respectively. Suppose that the arrivals and departures ofboth local and migrated VMs follow Poisson process mod-el. The system state transition is formulated as continuous-time Markov process. Let n𝑙 = (𝑛𝑙

1, ⋅ ⋅ ⋅ , 𝑛𝑙𝑘, ⋅ ⋅ ⋅ , 𝑛𝑙

𝐾) andn𝑔 = (𝑛𝑔

1, ⋅ ⋅ ⋅ , 𝑛𝑔𝑘, ⋅ ⋅ ⋅ , 𝑛𝑔

𝐾). The system state is representedby s = (n𝑙,n𝑔). Given arrival and departure rate of new andmigrated VMs, the steady state probability matrix Π will bederived by a 2𝐾-dimension Markov chain model.

Let 𝑅𝑏 and 𝑅𝑑 denote the blocking rate and the droppingrate, respectively. Then, a new local VM is blocked if the totalamount of required resources of local VMs (including the newone) exceeds that of common resources, i.e.,

∑𝐾𝑘=1 𝑛

𝑙𝑘𝑐𝑘 > 𝐶𝑐

or∑𝐾

𝑘=1 𝑛𝑙𝑘𝑚𝑘 > 𝑀𝑐. A migrated VM is dropped if the

total amount of required resources of all VMs (includingthe migrated one) is more than that of all resources, i.e.,∑𝐾

𝑘=1(𝑛𝑙𝑘 + 𝑛𝑔

𝑘)𝑐𝑘 > 𝐶 or∑𝐾

𝑘=1(𝑛𝑙𝑘 + 𝑛𝑔

𝑘)𝑚𝑘 > 𝑀 . Let 𝑅𝑐𝑏

denote the constraint of blocking rate. The optimal numberof reserved resources is derived by solving the followingoptimization problem.

min 𝑅𝑑(𝐶𝑟,𝑀𝑟);

s.t. 𝑅𝑏(𝐶𝑟,𝑀𝑟) ≤ 𝑅𝑐𝑏

(10)

544

Page 6: Virtual Machine Live Migration for Pervasive Services … Machine Live Migration for Pervasive Services in Cloud ... The application of mobile cloud computing ... namely, vehicular

Fig.4 shows the performance comparison with and withoutresource reservation. The total resources of roadside cloud are50 and 100 units in computation and storage, respectively.Two classes of VMs are considered. VMs of class-1 are forcomputation-type applications, which needs resources 20 unitsin computation and 15 units in storage. VMs of class-2 arefor storage-type applications, which need resources 10 unitsin computation and 40 units in storage. Results show that thedropping rate of migrated VM is significantly reduced withresource reservation, which demonstrates the efficiency of ourproposed mechanism to protect VM migration. In particular,the dropping rate could be decreased over 60% at low VMarrival rate and nearly 20% at high VM arrival rate.

V. CONCLUSION

The application of mobile cloud computing in traditionalVANET gives rise to the new paradigm of cloud-assistedvehicular networks. The high mobility of vehicles causes a sig-nificant challenges for providing uninterrupted cloud servicesin cloud-assisted vehicular networks. VM live migration is apreferred means to shift to cloud services from source cloudsite to target cloud site. To improve the efficiency of dirty pagetransfer in VM live migration, this paper proposes a selectivedirty page transfer strategy, which is shown to enhance thedirty page transfer rate by 3.2 times. In addition, to reducethe phenomena of migration dropping, an optimal resourcereservation scheme is designed, which is demonstrated todecrease the dropping rate up to 60%.

ACKNOWLEDAGEMENT

This research is partially supported by program of NSFC(grant no. U1035001, U1201253, 61203117), the OpeningProject of Key Lab. of Cognitive Radio and Information Pro-cessing (GUET), Ministry of Education (grant no. 2011KF06),the project 217006 funded by the Research Council of Norway,the European Commission FP7 Project EVANS (grant no.2010-269323), and the European Commission COST ActionIC0902, IC0905 and IC1004.

REFERENCES

[1] ITU Strategy and Policy Unit (SPU), ITU Internet Reports 2005: TheInternet of Things, Geneva, International Tele-communication Union(ITU), 2005.

[2] Yazhi Liu, Jianwei Niu, Jian Ma, Lei Shu, Takahiro Hara, “The Insightsof Message Delivery Delay in VANETs with a Bidirectional TrafficModel”. Journal of Network and Computer Applications, Elservier,2012.

[3] M. Miche, T. Bohnert, “The Internet of Vehicles or the Second Gener-ation of Telematic Services”, ERCIM News, vol.77, pp.43-45, 2009.

[4] R. A. Uzcategui, and G. Acosta-Marum, “WAVE: A Tutorial”, IEEECommunications Magazine, 47(5), pp. 126-133, May 2009.

[5] S. Xie, Y. Liu, Y. Zhang, and . Yu, “A Parallel Cooperative SpectrumSensing in Cognitive Radio Networks”, IEEE Transactions on VehicularTechnology, vol. 59, no. 8, pp.4079 C 4092, 2010.

[6] R. Yu, Y. Zhang, Y. Liu, S. Xie, L. Song and M. Guizani, “SecondaryUsers Cooperation in Cognitive Radio Networks: Balancing SensingAccuracy and Efficiency”, IEEE Wireless Communications Magazine,vol. 19, no. 2, April 2012, pp.2-9.

[7] T. Wang, L. Song, and Z. Han, “Coalitional Graph Games for PopularContent Distribution in Cognitive Radio VANETs,” to appear, IEEETransactions on Vehicular Technologies.

[8] T. Wang, L. Song, Z. Han, and B. Jiao “Popular Content Distributionin CR-VANETs with Joint Spectrum Sensing and Channel Access,” toappear, IEEE Journal on Selected Areas in Communications.

[9] R. Hussain, J. Son, H. Eun, S. Kim and H. Oh, “Rethinking VehicularCommunications: Merging VANET with cloud computing”, in Proc.IEEE 4th International Conference on Cloud Computing Technologyand Science, pp. 606-609, 2012.

[10] C. Clark, K. Fraser, S. Hand, J. G. Hansen, E. Jul, C. Limpach,I. Pratt, and A. Warfield, “Live migration of virtual machines,” in Proc.USENIX Symposium on Networked Systems Design and Implementation(NSDI05), Berkeley, CA, USA, 2005, pp. 273-286.

[11] P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, A. Ho, R. Neuge-bauer, I. Pratt and A. Warfield, “Xen and the Art of Virtualization,”in Proc. of the nineteenth ACM symposium on Operating systemsprinciples, pp 164-177, Bolton Landing, NY, USA, 2003.

[12] M. Rosenblum and T. Garfinkel, “virtual machine monitors: currenttechnology and future trends”, Computer, 38(5), pp. 39-47, May 2005.

[13] M. Mishra, A. Das, P. Kulkarni, A. Sahoo, “Dynamic resource man-agement using virtual machine migrations”, IEEE CommunicationsMagazine, vol.50, no.9, pp.34-40, 2012.

[14] S. K. Bose, S. Brock, R. Skeoch and S. Rao, “Cloudspider: Combiningreplication with scheduling for optimizing live migration of virtualmachines across wide area networks”, in Proc. International Symp. onCluster, Cloud and Grid Computing, pp. 13-22, 2011.

[15] K. Ye, X. Jiang, D. Huang, J. Chen, and B. Wang. “Live Migrationof Multiple Virtual Machines with Resource Reservation in CloudComputing Environments”, in Proc. IEEE CLOUD, July 2011.

[16] M. Satyanarayanan, P. Bahl, R. Caceres, and N. Davies. “The Case forVM-based Cloudlets in Mobile Computing”, IEEE Pervasive Comput-ing, 8(4), 2009.

545