The proposed algorithm selects a set of VMs in a VC to be migrated on satisfying the criteria of the provider such as minimum total migration time, little / minimum down time better power management and increase in performance etc.
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0 ratings0% found this document useful (0 votes)
110 views5 pages
Paper26 PDF
The proposed algorithm selects a set of VMs in a VC to be migrated on satisfying the criteria of the provider such as minimum total migration time, little / minimum down time better power management and increase in performance etc.
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5
International Conference on Computing and Intelligence Systems
Pages: 1252 1256
Volume: 04, Special Issue: March 2015
ISSN: 2278-2397
Analyzing a Novel Approach for Virtual
Cluster Migration in Cloud Computing T.Lavanya Suja1, V.Savithri2 1
MPhil Scholar, Dept.of.Computer Science, Mother Teresa Womens University, Kodaikanal, India 2 Assistant Professor, Dept. of Computer Science, Womens Christian College, Chennai, India
Abstract-Virtual Cluster (VC) migration is
needed in Cloud Computing (CC) during situations like load balancing, fault tolerance ,complying with Service Level Agreement(SLA) etc. While migrating VC from one Physical Machine(PM) to another PM various constraints are considered for an optimal solution. The proposed Backtracking algorithm which has provided optimal solutions for popular problems like N-Queens, Hamiltonian cycle, selects a set of VMs in a VC to be migrated on satisfying the criteria of the provider such as minimum total migration time, little/minimum down time better power management and increase in performance etc. Keywords - Migration; clusters; VM; VC; Backtracking; cloud computing I. INTRODUCTION The term Cloud Computing (CC)has wide spread and has taken many dimensions in IT industry. Even then we need a simple go through regarding its definition and working. It can be defined simply as a service which can be rented from internet on a pay per use basic just like electricity, gas, water etc.The services can be software,platform or an infrastructure.Thus the famous three services of CC are called as Software-as-a-Service(SaaS) Platform-as-a-Service(PaaS) Infrastructure-as-a-Service(IaaS) The Iaas working goes like this. The provider rent a machine(s) to the consumer based upon the request given by the consumer after specifying clearly it in Service Level Agreements(SLA).The consumer can develop and build applications for his convenience on that rented one(s). That machine is called a Virtual Machine(VM) and a group of VMs are called Virtual Cluster(VC) which resides in a Physical machine. PM can be visualized as a group of powerful configuration servers. During working or later at some point due to necessities such as load balancing, fault tolerance and complying with SLA the provider has to move VCs from one PM to another PM. The advantage of VM is that, the higher configuration server machine can be able to run on a minimum configuration user machine which is achieved via IaaS of cloud computing. The concept of grouping one or more VMs into VC saves time and gives more performance in VC migration.
International Journal of Computing Algorithm (IJCOA)
The rest of the paper is about IaaS and its working
alone. The scope of this paper is to survey the techniques, approaches, algorithm used for migration and to analyze their pro s and cons to bring out a new approach for VC migration. II. LITERATURE SURVEY Vikram Yadav, Pooja Malik and Ajay Singh Chauhan suggest an optimization for VM which is energy efficient by proposing a term Powernap. Powernap is a useful state which saves more power as only the system components which checks the job arrivals will be active while CPU and DRAM are inactive. Also they introduce a new component in the proposed system called Power Consumption Tuner node(PCTN) will be in sync with the controller node of the Existing system which incorporates Ant colony optimization for visiting all nodes in the cluster and Bee hive optimization for Load balancing. They have used past 3 years of data for Job arrival so as to use the proposed system to avoid Wake delay and Nap Delay by extending the previous job till the new job is allocated [1] Ajeena Beegom ,and M S Rajasree propose an algorithm for VM provisioning and VM termination on the basis of availability of resources. A centralized approach in which the Load Prediction Module takes the decision whether to create a new VM in an already existing PM or to create a VM in a new PM by consulting with the centralized controller module and a Database on VM.Their algorithm employs greedy method and claims it to be a cost optimal one with running time O(nk).[2] Another algorithm given by Harpreet Kaur and Maninder Singhis a sequential one and her each user requests are considered as a process. The processes are assigned by comparing the memory requirements of the process and the cloud. Process migration is done from one cloud to another if it is not completed within its deadline.[3] Sharmila S. Patil, and Priyanka Pattenshetti propose 'Green Cloud Computing' by offering 2 methods namely Static and Dynamic architecture energy management. However both methods achieve power saving at the cost of performance degradation of 3%.[4] As [5] introduces the concept of virtualization by combining two nodes to form a 2-node cluster.When any one of the node becomes faulty with no time and without the knowledge of the user the other node
1252
International Conference on Computing and Intelligence Systems
Pages: 1252 1256 comes into operation. The Heartbeat Apache is used to configure 2-node cluster.Their results showed a reliable,scalable and high performance situation due to virtualization. Also it is client side compatible,cost effective and little or no overhead . Poral Nagaraja and S.Tejaswini.have used a 'Green algorithm' which migrates VM or turn off PM by comparing the resource usage with 'Hot Threshold' and 'Cold Threshold'. Also they have introduced a 'skewness' concept to decrease the irregularity of VM allocation so as to achieve Green Computing. Simulation results using Cloud SIM shows that overload avoidance mapping of virtual to physical resources adaptively based on changing requirements are achieved. [6] Comparison of two algorithms namely 'Time-shared' and 'Space-shared' approached on the 1000 cloudlets(tasks) and inference that on the basis of Task Profit the Space-shared one outdid the Time shared approach by completing 97% of the tasks before deadline. Although there is only marginal difference in the output on Task Profit,when the parameters such as Throughput,Task Penalty and Net Gain are considered the Space-shared approach outweighs the Time-shared approach. It is very interesting to know the result that space shared is the better approach for tasks scheduling in cloud environment.[7] A study was conducted to run the two codes of NQueens problem and Towers of Hanoi in different Virtual machines. Their individual execution time (in seconds) is measured and compared. Out of the 3 VMs RedHat Linux VM gave the least execution time. This paper serve as an introduction for Virtualization concept and comparison of their characteristics which is very primary notion in cloud computing. This knowledge can be applied in larger scale real time situations in CC so as to select VMs on the basis of specific performance criteria. [8] Discussion about the working of various algorithms that are used for soft computing and analyze the fact which one can be best suitable of Load Balancing in cloud computing. The load balancing can be CPU ,Memory, Performance,Delay etc. Among the 8 algorithms analyzed Active clustering and Ant colony are potential candidates for cloud computing.[9] Supreet Kaur Sahi et al. analyze the already existing resource prediction models and try to find out a better approach for resource prediction which can be well suited for E-business applications. They have also provided the statistics that every 100ms delay in Amazon costs a reduction of 1% sales. Another data, points out that a 43 billion dollar loss was faced due to bad performance of the e-commerce websites [10].This proves the fact that Delay time or Down time is a crucial factor in designing algorithms for VC migration and it should be very little if needed. In [11] the authors introduce the concept of Virtual Cluster(VC) which is a group of VMs in a PM.Homogenous and Heterogeneous VC migration
International Journal of Computing Algorithm (IJCOA)
Volume: 04, Special Issue: March 2015
ISSN: 2278-2397
are performed in a simulated environment using
Sequential,Concurrent, Mutual migration techniques.The results are very useful and interesting to device any algorithm in future. They suggest that if there is a master and slave VC then slave VC should be given high priority for migration. Also their findings that the first VC that was migrated gave better performance than the later migrated VCs. Andreas Berl et al.find prospect of energy efficiency in cloud computing by bringing out the necessity of it. Heating caused by IT industry leads to negative impact of Green House Gases (GHG) by excessive Co2 emission. Thus the authors bring out the need of "Smartness" in "Computing" by being "energy efficient". Virtualization is a potential technology which saves energy for cloud computing. VMs can be migrated from one PM to another PM so as to reduce the heat generated by more PMs.Some PMs can be switched off if there is no VM running on it. The suggestions such as as Full and Para Virtualization,Dynamic Resource Provisioning are implemented for energy efficient cloud computing.[12] Another paper discusses the VM live migration while maintain transparency in the application,o.s and remote client level.Their results show that the VM experiences a down time of only less than a second which provides transparency to the user.The total migration time can be controlled by properly managing the CPU resources. This method of using VMWare differs from a previous study by Hansen and Jul as it migrates a VM unmodified whereas their method needs some modification in the guest o.s.[13] Anton Belaglosav and Rajkumar Buyya portrays that their algorithm, is suited for generic cloud environment, ly to adheres strictly to SLAs and gives better performance. Their simulation results showed that it is possible to device an algorithm on 3 constraints such as resources,energy and thermal optimization which achieves a better performance for a generic cloud environment with dynamic workloads.[14] Cloudsim can be effectively used as simulation toolkit for studying,analyzing,simulating the behaviours of cloud systems. As the cloud architecture is enormous and time,cost consuming for implementation before knowing the results, this toolkit serve as a helpful alternative for researchers. The authors have done case studies of different scenarios like, scalability and overhead evaluation for federated and non-federated clouds where federation clouds did well, computed the makespan of hybrid cloud strategy where public cloud inclusion gave good cost returns,the last case study was related to energy conscious management of datacenter which revealed in order to save energy some VMs are migrated that caused an average of 70% SLA violations which needs to be improved.[15] The results give us a caution that SLA violations should not be occurring as a result of VM migrations. A new algorithm for Load Balancing includes two factors, Load and Time
1253
International Conference on Computing and Intelligence Systems
Pages: 1252 1256 as the criteria. As per the algorithm when the load of a Data center goes above a Threshold value then its load will be transferred to any other data center which is 5 and 1/2 hours geographically apart from it. The approach is considered feasible as the time zone changes; the latter data center will have fewer loads than the former. [16] This provides an option for selecting PMs for migration. The leasing concept in cloud using 'Haizea' as backend in OpenNebula. There are 3 approaches available in it namely Advance Reservation,Immediate and Best Effort of which the first two are given higher priority than the third one.[17] Various techniques such as having a Monitoring Agent,Threshold Value,assigning priority,using Global Polls,social cloud,Queue based model,Simulated Annealing model,Energy conserving model and uncertainty game theory as the logic for Resource allocation algorithms. The paper concludes with a remark that cloud will be the 5th utility such as water,electricity etc, so it is a must to manage its resources to maximize the profit of provider and minimize the cost of the consumer. [18] The data leakage problem in cloud computing is discussed as data is an important asset that has to be secured always. Semantic Clustering technique is applied for data thereby anomalies are detected and threats are found out followed by database updation leading to its security.[19] Comparison of the solutions provided by Backtracking and Genetic algorithms for the NQueens problem. The author concludes with the fact after several iterations performed on both of them,that Genetic algorithm provide simpler solution but the execution time of Backtracking algorithm is very less than the former so the efficiency of backtracking is proved.[20] A methodology is suggested where a unique solution for N-Queen problem is obtained. There are two types of solutions for a problem. Theyare Unique and Distinct. In that case there are a total of 92 solutions out of which only 12 are different solutions . The authors bring out 3 rules in order to find a unique solution.Their unique solution holds for all values n>4. [21] Somaeyh Soltan Bagshahi et al.propose Greedy Algorithm for VC migration.The only criterion taken was the size of the VM, PM and no other constraints are considered.Their approach of taking a Virtual cluster as a combination of several VMs is useful and will save sometime. Their simulation results proved that Greedy Algorithm gave better performance than the serial method.[22] A survey of various live migration techniques and does a comparative study on them.They try to analyze the algorithms of live migration ,find out that all of them attempt to reduce down time and increase the performance than before.The authors find out that PreCopy approach interesting and live VM migration for Cloud Federation to be promising in the future.As a
International Journal of Computing Algorithm (IJCOA)
Volume: 04, Special Issue: March 2015
ISSN: 2278-2397
result we get to know what are the various criteria on
which approach should be proposed such as power management,proactive maintenance and Load balancing etc.[23] Pallavi Gupta et al.propose a power-efficient VM migration. Their approach consolidates the VMs so as to be energy saving.The output of the method achieves mapping of several VMs to fewer PMs and switching off the underutilized or idle PM there by saving energy and produces less heat by leaving the minimum carbon foot print.Their achievement in power saving results contribute to Green computing as well.[24]. Gulshan Soni and Mala Kalra introduce the dimension of security in Live VM migration.They point out the fact where there are some situations in which the security of VMs (source and destination) and hypervisor can be at risk by intruders and malfunctioning. The facts like Data confidentiality, Data Integrity and Privacy comes into picture during security breach.They analyze few solutions such as Hash code,security to system calls between Hypervisor and guest o.s etc.Since the data transferred between VMs are plain text and they go through a network, security is also a mandatory factor to be considered.[25] In [26] theauthors survey the cloud computing providers and tabulate their services and turnovers. Also some important research groups and their works are tabulated .A study on the impact of cloud computing revealed that 54% of CIOs have evaluated cloud computing and 46% of CIOs have yet to cloud computing. The authors suggest that the cloud migration can be supported by a Cloud Adoption ToolKit which helps in the analysis of decision. Pradip et al. justify the fact that live migration is very essential in VM migration. Two categories of it are Pre-copy and Post-copy are analyzed and comes out with the fact the pre-copy is the most suitable one for live VM migration. The authors surveyed 7 research papers for VM migration and formulated 6 Performance Metrics such as Preparation time, Down time, Pages transferred, Resume time, Total migration n time and Application degradation. Lastly they include security is also an important factor in VM migration.[27].A case study details the portability between Eucalyptus and Open Stack. The inference got from the case study was that portability between different cloud providers. Also it will attract more users to shift into cloud computing. Standardization is the need of the hour which the author tries to bring out in this paper.[28] Risk Homeostasis Methodology where in which a model is used to analyze the cloud computing alternative for the underlying business system.This model first assesses the risk and analyzes thebenefits. Then a migration strategy is designed.Next step is to migrate and manage by Change management.The last step is evaluation by doing Business Impact and Stakeholder evaluation.[29] A case study conducted by Ali Khajeh et al. reveals when migrating on-premises applications
1254
International Conference on Computing and Intelligence Systems
Pages: 1252 1256 to cloud data center the organization has to undergo a transition phase which will be revealing the positive of the cloud while sacrificing the well accustomed usage of Legacy applications even though they enjoy benefits of migration and propose a method whichmitigates it for any type of organization that likes to adopt cloud computing.[30] III. FINDINGS After analyzing 30 research papers we are able to find out the fact that there is a need for an effective algorithm which accomplishes live migration without compromising SLA agreements. One of the authors [22] proposes some famous algorithm like greedy approach aiming for optimal solution. This gave us an idea that Backtracking can be a potential candidate for VC migration algorithm. Backtracking is a popular algorithm formulated by B.H.Lehmer in 1952. By correlating backtracking with VC migration the method proposed can be given as 1. Aim Migrate VCs from one PM to another PM based upon the constraints. 2. Explicit Constraints Select VMs from the same VCs 3. Implicit Constraints 4. Memory size of VM and PM Minimum down time Increase in performance after migration 5. Pseudocode Procedure VC_Migration( ) Begin vclist, pmlist arranged in descending order; VMs arranged in ascending order of downtime; For each PM in pmlist do While sizeof(PM) > 0 do For each VC in vclist do If sizeof(VC)> sizeof(PM) then continue; For each VM in VC do sizeof(PM) = sizeof(PM) sizeof(VM) End for VC is in migration to PM End for End while End for End IV. CONCLUSIONAND FUTURE SCOPE Conducting a wide survey about VC migration and its techniques, we have found out that there is a need for better algorithms for VC migration which achieves energy efficiency. This will attract more avenues of applications for CC and also pave way for Green Computing. The analysis of various algorithms proved the fact that the proposed one can be better than the existing ones. Though in paper [22] the authors introduced a new technique called GreedyMethod
International Journal of Computing Algorithm (IJCOA)
Volume: 04, Special Issue: March 2015
ISSN: 2278-2397
and proved it to be better than the Serial Method,it
failed to include several important constraints like minimum downtime,better performance etc. Apopular algorithm Backtracking[31] was found as a potential candidate for this and a rough pseudo code is developed.In contrary to the Greedy Method this algorithm incorporates constraints like minimum downtime,selecting VMs from same VCs, hence better performance will be achieved than the former methods like Serial and Greedy approaches. Future work includes fine tuning the code for the algorithm to include more implicit constraints and studying the results. We propose to do this by CloudSimsimulation software and come up with the simulation results in future. REFERENCES [1]
Vikram Yadav, Pooja Malik, Ajay Singh Chauhan, Energy
efficient VM optimization, International Journal of Computer Applications Volume 106 No.7, November 2014,pp. 23-28. [2] Ajeena Beegom , M S Rajasree ,Resource provisioning and management for IaaS providers in cloud computing, International Journal of Computer Applications (0975 8887) Volume 104 No.17, October 2014,pp. 1-4. [3] Harpreet Kaur,ManinderSingh, A Task scheduling and resource allocation algorithm for cloud using live migration and priorities, International Journal of Computer Applications (0975 8887),Volume 84 No 13, December 2013,pp. 7-12. [4] Sharmila S. Patil, Priyanka Pattenshetti, Overview of green cloud architecture, International Journal of Computer Applications (IJCA) (0975 8887) National Seminar on Recent Trends in Cloud Computing, NSRCC-2014,pp. 9-12. [5] Mamta Gadoria,Sarvesh Singh,Pratibha Expedient approach for high availability in web server services for HPC attained by clustering using virtualization, International Journal of Computer Applications (0975 8887)Volume 95 No.20, June 2014,pp.21-24. [6] Poral Nagaraja,Tejaswini.S. Autonomous management System for dynamic resource allocation in cloud computing, InternationalJournal of Computer Applications (0975 8887) International Conference on Information and Communication Technologies (ICICT- October 2014),pp.16-21. [7] Himani,Harmanbir Singh Sidhu,Comparative analysis of scheduling algorithms of cloudsim in cloud computing, International Journal of Computer Applications (0975 8887)Volume 97 No.16, July 2014,pp.29-33. [8] Vikas Verma,Sandeep Ranjan,Yash Pal,A Comparative study of open source operating systems for virtualization with executing applications,International Journal of Computer Applications (0975 8887)Volume 94 No.7, May 2014,pp.912. [9] Mayur S Pilavare,Amish Desai,A Survey of soft computing techniques based load balancing in cloud computing,International Journal of Computer Applications (0975 8887) Volume 110 No. 14, January 2015,pp.22-25. [10] Supreet Kaur Sahi,V.S.Dhaka,A Review on workload prediction of cloud services,International Journal of Computer Applications (0975 8887) Volume 109 No. 9, January 2015,pp.1-4. [11] Kejiang Ye, Xiaohong Jiang, Ran Ma, Fengxi Yan, VCMigration: Live Migration of Virtual Clusters in the Cloud, ACM/IEEE 13th International Conference on Grid Computing September 2012,pp.209-218. [12] Andreas Berl, Erol Gelenbe, Marco di Girolamo, Giovanni Giuliani,Hermann de Meer, Minh Quan Dang, Kostas Pentikousis,Energy-efficient cloud computing, The Computer Journal Advance Access published August 19, 2009,pp.1045-1051.
1255
International Conference on Computing and Intelligence Systems
Pages: 1252 1256 [13] Michael Nelson, Beng-Hong Lim,Greg Hutchins,Fast transparent migration for virtual machines,,VMware, Inc.Palo Alto, CA 94304,pp 25-25. [14] Anton Belaglosav,Rajkumar Buyya, Energy efficient resource management in virtualized cloud data centers, 10th IEEE/ ACM International Conference on Cluster,Grid and Cloud Computing,2010,pp.826-831. [15] Rodrigo,Rajiv,Anton,Cesar,Rajkumar Buyya,CloudSim a toolkit for modeling and simulation of cloud computing environments and resource provisioning algorithms,wileyonlinelibrary.com,published online 24 Aug 2010,pp.23-50. [16] Parampreet Singh Jaura, Time and load based cloud Scheduling algorithm,International Journal of Computer Applications (0975 8887) Volume 110 No. 11, January 2015,pp.34-40/ [17] Avanika M Chavada,Upendra R Bhoi,Review on scheduling of resources in cloud environment, International Journal of Computer Applications (0975 8887)Volume 110 No. 4, January 2015,pp.16-18. [18] Sagar S. Brahmkshatriya,Jignesh Prajapati, A Review on techniques of resource allocation in cloud computing, International Journal of Computer Applications (0975 8887) Volume 110 No. 4, January 2015,pp.13-15. [19] C.Suresh Kumar, K. Iyakutty, Semantic cluster based classification for data leakage detection for the cloud security, International Journal of Computer Applications (0975 8887)Volume 110 No. 6, January 2015,pp.19-22. [20] VikasThada, Shivali Dhaka, Performance analysis of NQueen problem using backtracking and genetic algorithm techniques, International Journal of Computer Applications (0975 8887)Volume 102 No.7, September 2014,pp.26-28. [21] [21]Vishal Kesr,iVaibhav Kesri, An unique solution for N queen problem, International Journal of Computer Applications (0975 8887)Volume 43 No.12, April 2012,pp.16. [22] Somayeh Soltan Baghshahi,Sam Jabbehdari,Sahar dabi,VM migration based on greedy algorithm in cloud computing,International Journal of Computer Applications (0975 8887) Volume 96 No.12, June 2014,pp.32-35. [23] Gulshan Soni, Mala Kalra, Comparative study of live VM migration techniques in cloud. International Journal of Computer Applications (0975 8887) Volume 84 No 14, December 2013,pp.19-25. [24] Pallavi Gupta,Lokendra Vishwakarma,Awadheshwari Patel, Power aware VM consolidation considering multiple recourses with live migration. International Journal of Computer Applications (0975 8887)Volume 103 No 17, October 2014,pp.24-30. [25] Suresh B.Rathod, V Krishna Reddy,Secure live VM migration in cloud computing : A Survey. International Journal of Computer Applications (0975 8887) Volume 103 No.2, October 2014,pp.18-22. [26] Neetika Gupta,Jyoti Kataria,Abhay Bansal, A Survey on cloud providers and migration issues. International Journal of Computer Applications (0975 8887)Volume 56 No.14, October 2012,pp.38-43. [27] Pradip D. Patel,Miren Karamta,M. D. Bhavsar, M.B.Potdar,Live VM migration techniques in cloud computing: A Survey., International Journal of Computer Applications (0975 8887)Volume 86 No 16, January 2014,pp.18-21. [28] Pushpanjali Arora,Kashyap,MaitreyeeDutta, Portability between Eucalyptus and Open stack : A Case study of the challenges involved in cloud Migration. International Journal of Computer Applications (0975 8887)International conference on Green Computing and Technology, October 2013,pp.23-25. [29] HamseYuusuf,ChristopherTubb, Migration to cloud computing: A risk homeostatismethodology,International Journal of Computer Applications (0975 8887)Volume 77 No.7, September 2013,pp.5-9. [30] Ali Khajeh Hosseini,David Greenwood Ian Somerville, Cloud Migration: A Case Study of migration an enterprise IT system to IaaS, IEEE 3rd International conference on
International Journal of Computing Algorithm (IJCOA)
Volume: 04, Special Issue: March 2015
ISSN: 2278-2397
cloud computing 2010,pp.450-457.
[31] EllisHorrowitz,SartajSahni, Fundamentals of Algorithms ,Computer Science Press, pp. 323-350.