IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 277-284

International Journal of Research in Information Technology (IJRIT) www.ijrit.com

ISSN 2001-5569

Improving Resource Utilization by Minimum Distance Based Virtual Machine Placement Algorithm Chirag Rathod1, Mohammed Hussain 2 1

Student of Master in Computer Science & Engineering Gujarat Technological University, Ahmedabad,Gujarat. [email protected] 2

Asst. Prof,CSE Dept PIET Vadodara,Gujarat. [email protected]

Abstract Cloud computing provides platform by which user meet their demand in efficient way. Cloud computing used virtualization technologies for getting maximum utilization of available resources. Virtual Machines are utilized to satisfy user requirement and placed on Physical machine of cloud for Effective usage of available hardware resources and electricity used in the cloud. Reducing the number of running physical machine helps in cutting down power consumption in cloud. An efficient technique is to place virtual machine in to physical machine such that utilization of available physical machine goes high and required number of physical machines are minimize. Due to multiple dimensionality of physical resources, waste of resources by imbalanced use of multidimensional resources. To describe multi-dimensional resource usage states of physical machines, multidimensional normalized resource cube is represented. According to this model we propose “ A minimum distance based virtual machine algorithm” which can balance multi-dimensional resources. Thus utilization of available physical machines are increase and number of running physical machines are minimize. We also evaluate our proposed algorithm using Netbeans and experiments on Cloudsim Toolkit.

Keywords: VM, PM, TCP, TUP, RDV, Multi Dimensional resource.

1. Introduction A cloud is a large pool, of easily and accessible virtualized resources, such as hardware, development platforms and/or services. These resources can be powerfully re-configured to arrange properly to a variable load scale, and also permitting for an optimum resource use. This pool of resources is constituting a type exploited by a pay-per-use model in which guarantees are hold out for acceptance by the infrastructure supply by means of usage Service -Level Agreements (SLA). [1] In the cloud, the end user is just using a very light device which is capable of using a network that connects it to a server at some other location. The users do not need to store the data at its end as all the data is stored on the remote server at some other place. A cloud is a pattern of parallel and distributed system be composed of a collection of interconnected and virtualized computers that are dynamically stipulation and presented as one or more unite computing resources established on service level agreements found amongst negotiation between the service supplier and consumer. It uses remote services through a network using various resources. It is basically meant to give maximum with the minimum resources i.e. the user end is having the minimum Chirag Rathod,IJRIT

277

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 277-284

hardware requirement but is using the maximum capability of computing. This is possible only through this technology which requires and utilizes its resources in the best way. The concept of cloud computing is linked closely with those of Information as a service (IaaS), Platform as a service (PaaS), Software as service (SaaS) all of which means a service-oriented architecture. Here comes the first benefit of the cloud computing i.e. it reduces cost of hardware that could have been used at user end. As there is no need to store data at user's end because it is already at some other location. So instead of buying the whole infrastructure required to run the processes and save bulk of data you are just renting the assets according to your requirement. The similar idea is behind all cloud networks. Virtualization creates a virtual version of resources and share among multiple users using hypervisor (XAN [4], KVM, VMware) by sharing resources among multiple user virtualization increase resource utilization and reducing number of hardware requirement. However there may be a number of underutilized physical machines due to the inefficient mapping of virtual machines to physical machines.

2. Virtual Machine Placement The problem of virtual machine placement is core of cloud computing. Virtual Machine Placement [6] is a mapping between physical machines to virtual machines. Better placement strategy is to decrease running physical machines, increase the performance of infrastructure.

Fig 1 Work flow of Cloud system and VM placement. [1] Cloud computing refers to the delivery of computing resources over the Internet.[1] Instead of keeping data on your own hard drive or updating applications for your needs, you use a service over the Internet, at another location, to store your information or use its applications. whenever user request for any resource its considered as task in fig 3.1. User request placed in to virtual machine through Datacenter Broker layer. virtual machine is an abstraction of physical machine. it will acts like an real machine. it is giving with memory cpu and bandwidth from underlying physical host. Multiple VM can run on single physical machine if host have enough capacity. Virtual machine place in to physical machine according to VmAllocation or VmPlacement policy. VmPlacement policies decided that Which VM allocate to which host such that utilization host increase. Based on placement goal, algorithms will be divided into two categories.

Chirag Rathod,IJRIT

278

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 277-284

• •

QOS based Algorithms [8]: By providing maximum resources to virtual machines, then achieving the maximum quality of service. Maximizing the quality of a service in any situation is the goal of these algorithms. Power Based [8]: Power might be save with minimize number of running physical machines. Minimize number of running physical machines without exceeding the probability of violating the user service level agreements.

3. Virtual machine placement algorithms Cluster controller will collect resource utilization information from all node controllers with in the cluster. Using collected information, cluster controller will find mappings between virtual machines and node controller (physical machines). 1. First Fit [9]: While placing the VMs, first fit algorithm can be seen as a locally optimal algorithm in greedy manner. When a VM request arrives, the firstly scanned one with sufficient available resources will be selected to host the VM. To avoid unnecessary scanning, the fully loaded PMs are not scanned. If there is no PM with sufficient available resources, a new PM will start up and create the requested VM on the newly started PM. It is a greedy manner to place the VMs and, by this way, we can achieve a local optimization. Physical machines are available with their indices. Each virtual machine will start searching from starting node, whenever the physical machine fulfills requirement of the virtual machine, then virtual machine will start on that node. Average time complexity of this algorithm is O(n/2). 2. Next Fit [9]: Algorithm will use variable name called next. At Initial, next is null. First virtual machine search starts from the first node. When physical machine satisfies the resource requirement, virtual machine starts on that physical machine and next is replaced by current physical machine. If there is no PM with sufficient available resources, a new PM will start up and create the requested VM on the newly started PM. For the next virtual machine search starts from next to the stored next physical machine, search continues up to desired physical machine. 3. Random Fit [9]: Randomly a physical machine is chosen for placing virtual machine. If physical machine satisfies the resource requirement, virtual machine starts on that physical machine. If it not satisfied randomly another physical machine is chosen. This process will continue for every virtual machine. 4. Least full first[10]: Find the physical machine which is least full. If the physical machine satisfy resource requirement of virtual machine then virtual machine will start. If it doesn’t satisfy other physical machines also doesn’t satisfy. Because virtual machine is been tested with least full physical machine. This process will continue till virtual machines exhausted. Time complexity of this algorithm is O(n). 5. Most Full First[10] : In this process, physical machines are sorted from most full to least full. Once sorted has been done first fit process will takes place. This process continues up to virtual machine allocated to any host. Time complexity of this algorithm is O(n). 6. First Fit Decreasing (Single Dimension)[11] : Both physical machines and virtual machines are sorted from high capacity to least capacity machines. Once both physical and virtual machines are arranged, First fit process takes place. Time complexity of this algorithm is O(n+m) 7. First Fit Decreasing product[12]: This is also similar to the First Fit Decreasing. Each physical machine or virtual machine is represented with different resource capacities and each resource capacity is normalized in to percentage of utilization. Example: If physical machine utilization of RAM is 512Mb and total RAM is 1024Mb then RAM utilization is 50%. Volume of physical machine or virtual machine can be defined as product of all resource types. This volume is used instead of capacity in First Fit Decreasing (Single Dimension). 8. First Fit Decreasing sum[5]: This is also similar to the First Fit Decreasing. Each physical machine or virtual machine is represented with different resource capacities and each resource capacity is normalized in to percentage of utilization. Virtual or physical machines volume is defined as weighted sum of values. i.e. V = sum(wi _ Ri), where i th resource weight wi= total demand for i th resource/total available capacity for i th resource. This volume is used instead of capacity in First Fit Decreasing(single Dimension). 9. Minimum Magnitude Based VM placement[5] : All physical machines resource utilization vector (RUVi) are known and also virtual machine resource demand is known. For each physical machine, if it satisfy the resources demand of virtual machine then find the angle between RUVi + RDV vector and total capacitance vector TCV (1,1). In figure- 2.1 the angle shown by c. Place the virtual machine on which will give minimum angle. Chirag Rathod,IJRIT

279

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 277-284

Fig 2 Minimizing Angle virtual machine placement [5].

4. PROPOSED WORK Problem With Minimizing angle is this algorithm utilize the resource but increase the number of physical machine. This algorithm choose the minimum angle physical machine to place virtual machine. If less utilized PM shows less angle than the high utilized PM than its place VM to less utilized PM. To overcome the problem of Angle based we proposed “Minimum Distance based virtual machine algorithm”. Here we calculate the distance between Temporal resource Utilization Point (TUP) and Total Capacity Point (TCP) of the host before placing the virtual machine in to host. See Fig 3

Fig 3 Minimum Distance based VM Placement In Minimum distance based algorithm We calculate Distance Between TUP & TCP for each and every host which has enough capacity to satisfy requested VM. After calculating the Distance (TUP,TCP) we find out the host with minimum Distance (TUP,TCP) and Place VM to PM which has lowest Distance(TUP,TCP). By Minimum Distance VM Placement Algorithm we can utilize resources in each dimension (Here we take

Chirag Rathod,IJRIT

280

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 277-284

just Two Dimensional Host). By efficient use of hardware number of running physical machines are reduce. 1. PMi=(RCi1,RCi2,….RCp) //Physical machine capacities , Toatal P PM available 2. Vj=(RDj1,RDj2…….. RDjv)//VM resource Demand Total V number of VMs are available 3. While All Unplace VMS do 4. Num VMs
Num VMs+1 5. Calculate PMRi ( RRi1, RRi2, …..RRid) // Reserved resources for physical machine. 6. RPMi = PMi – PMRi // Remaining capacity of Physical Machine 7. Find RCPi = RPMi / PMi // Remaining capacity point 8. If(RPMi > Vk) // Check Every PM to check that PM can host this VM, Vk is new VM. 9. For Each PM TRU = RUi + Vk //RUi is PM utilization before placing Vk, Vk is New VM. 10. Calculate temporal PM utilization point TRPi = TRUi / PMi. 11. For each host Calculate Distance ( TUPi, RCPi ) 12. Place VM to PM which has lowest Distance (TUPi, RCPi). 13. Else Introduce new PM. 5.

EXPERIMENTAL RESULT We evaluate our algorithm in Cloudsim 3.0.3, Using netbeans IDE 6.9.1. In cloudsim3.0.3 toolkit we create VM allocation policy according to existing as well as our proposed algorithm. For testing our methodology by we created different sized physical machines and random sized Virtual machines and run both algorithms multiple times. For displaying result comparison here we generate utilization graph for both the direction RAM & MIPS. Table I shows result comparison of random sized 10 hosts & 25 Vms. Table I: Result Comparison 10 PM & 25 VM Host ID Angle based Vm placement Distance based Vm placement RAM MIPS RAM MIPS 0 18.46 30.91 38.28 87.36 1 25.28 37.59 48.04 94.02 2 0 0 42.70 92.68 3 6.25 14.1 44.99 93.47 4 0 0 38.43 81.42 5 47.75 81.77 0 0 6 0 0 0 0 7 0 0 0 0 8 0 0 0 0 9 51.56 84.59 0 0

Chirag Rathod,IJRIT

281

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 277-284

Fig 4 Result Comparison Graph

After experiment multiple times we can compare number of running physical host for different configuration Table 2 shows number of running host in different configuration.

Total Host

10 10 20 25 35

Chirag Rathod,IJRIT

Table II: Result Comparison on number of running host. Number of VM Number of Number of running PM running PM Distance based Angle based Placement Placement 25 6 5 40 9 7 60 13 11 100 19 18 120 25 22

282

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 277-284

CONCLUSION After studying different VM placement algorithm, there are wastage of resources in multidimensional resources. we studying Angle based VM placement algorithm it utilize resources in each direction such as CPU,RAM etc. During execution of angle based and analysis of result we have identified the situation where it increases multiple running machine that cause high energy consumption in cloud data center. To overcome the utilization problem we have proposed Minimum distance based VM placement algorithm. The proposed algorithm tested with angle based algorithm by introducing user defined parameter in cloudsim tool kit. During the simulation we have identified the proposed algorithm utilize more number of host and reducing number of running host which is proportional to energy consumption in Cloud datacenter.

ACKNOWLEDGMENT We are highly indebted to Asst. Prof. Narender Reddy, for his valuable guidance and supervision regarding our topic as well as for providing necessary information regarding review paper. We are very much thankful to Prof. G. B. Jethava for helping me in text preparation.

REFERENCES [1].

[2]. [3].

[4]. [5].

[6].

[7].

[8].

[9].

[10]. [11].

[12]. [13]. [14]. [15].

R. Buyya, C. S. Yeo, S. Venugopal, J. Broberg, and I. Brandic, “Cloud computing and emerging it platforms: Vision, hype, and reality for delivering computing as the 5th utility,” Future Gener. Comput. Syst., vol. 25, no. 6, pp. 599–616, Jun. 2009. O. D, “Cloud computing service models , part1: Infrastructire as a service,” CEO, Creative RIA, Jan. 2011, doi:10.1109/COASE.2010.5584153. H. La and S. Kim, “A Systematic Process for Developing High Quality SaaS Cloud Services,” in Cloud Computing, ser. Lecture Notes in Computer Science, M. Jaatun, G. Zhao, and C. Rong, Eds. Berlin, Heidelberg: Springer Berlin / Heidelberg, 2009, vol. 5931, ch. 25, pp. 278–289. C. C. U. C. Group, “Cloud computing use cases white paper version 1.0,” http://www.scribd.com/doc/18172802/Cloud-Computing-Use-Cases-Whitepaper, 2009. S. H. J. G. H. E. J. C. L. I. P. A. W. Christopher Clark, Keir Fraser, “Live migration of virtual machines,” in Proceeding NSDI’05 Proceedings of the 2nd conference on Symposium on Networked Systems Design & Implementation, vol. 2. USENIX Association Berkeley, CA, USA 2005, 2007, pp. 273–286. P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, A. Ho, R. Neugebauer, I. Pratt, and A. Warfield, “Xen and the art of virtualization,” in Proceedings of the nineteenth ACM symposium on Operating systems principles, ser. SOSP ’03. New York, NY, USA: ACM, 2003, pp. 164–177. R. Moreno-Vozmediano, R. S. Montero, and I. M. Llorente, “Iaas cloud architecture: From virtualized datacenters to federated cloud infrastructures,” Computer, vol. 45, no. 12, pp. 65–72, 2012.34. K. Mills, J. J. Filliben, and C. Dabrowski, “Comparing vm-placement algorithms for on-demand clouds.” in CloudCom, C. Lambrinoudakis, P. Rizomiliotis, and T. W. Wlodarczyk, Eds. IEEE, 2011, pp. 91–98. S. Shang, B. Wang, J. Jiang, Y. Wu, and W. Zheng, “An intelligent capacity planning model for cloud market,” Journal of Internet Services and Information Security (JISIS), vol. 1, no. 1, pp. 37– 45, 5 2011. A. Mehta, M. Menaria, S. Dangi, and S. Rao, “Energy conservation in cloud infrastructures,” IEEE SysCon 2011, vol. 1, pp. 456–460, 2011. J. Shetty, A. M. R, and S. G, “Article: A survey on techniques of secure live migration of virtual machine,” International Journal of Computer Applications, vol. 39, no. 12, pp. 34–39, February 2012, published by Foundation of Computer Science, New York, USA. R. G. Chris Hyser, Bret Mckee and B. J. Watson, “Autonomic virtual machine placement in the data center,” in HP Labs technical report, 2007. U. Bellur, C. S. Rao, and M. K. SD, “Optimal placement algorithms for virtual machines,” CoRR, vol. abs/1011.5064, 2010. A. Shankar and U. Bellur, “Virtual machine placment in computing clouds,” CoRR, vol. abs/1011.5064, April 10, 2010. W. F. de la Vega and G. S. Lueker, “Bin packing can be solved within 1+epsilon in linear time.” Combinatorica, vol. 1, no. 4, pp. 349–355, 1981.

Chirag Rathod,IJRIT

283

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 277-284

[16]. [17].

[18].

[19].

[20].

[21]. [22]. [23]. [24].

[25].

[26]

V. P. V. R. K. T. L. U. S. Lee, R. Panigrahy and U. Wieder, “Validating heuristics for virtual machine consolidation.” in Microsoft Research. Microsoft research, p. 14 pp. E. G. Coffman, Jr., M. R. Garey, and D. S. Johnson, “Approximation algorithms for np-hard problems,” D. S. Hochbaum, Ed. Boston, MA, USA: PWS Publishing Co.1997, ch. Approximation algorithms for bin packing: a survey, pp. 46–93. S. Srikantaiah, A. Kansal, and F. Zhao, “Energy aware consolidation for cloud computing,” in Proceedings of the 2008 conference on Power aware computing and systems, ser. HotPower’08. Berkeley, CA, USA: USENIX Association, 2008, pp. 10–10. 35. M. Mishra and A. Sahoo, “On theory of vm placement: Anomalies in existing methodologies and their mitigation using a novel vector based approach,” in Proceedings of the 2011 IEEE 4th International Conference on Cloud Computing, ser. CLOUD ’11. Washington, DC, USA: IEEE Computer Society, 2011, pp. 275–282. N. Singh and S. Rao, “Energy optimization policies for server clusters,” in 6th Annual IEEE Conference on Automation Science and Engineering (IEEE CASE 2010), Toronto, Canada, Aug. 2010, pp. 293–300, doi:10.1109/COASE.2010.5584153. A. Silberschatz, P. B. Galvin, and G. Gagne, Applied operating system concepts, 1st ed. New York [u.a.]: Wiley, 2000. B. Radojevic and M. Zagar, “Analysis of issues with load balancing algorithms in hosted (cloud) environments.” in MIPRO. IEEE, 2011, pp. 416–420. A. K. Norman Bobro and K. A. Beaty, “Dynamic placement of virtual machines for managing sla violations,” in In Integrated Network Management. IEEE, 2007, p. 119128. A. Kochut and K. Beaty, “On strategies for dynamic resource management in virtualized server environments,” in Proceedings of the 2007 15th International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems, ser. MASCOTS ’07. Washington, DC, USA: IEEE Computer Society, 2007, pp. 193–200. S. Shish and C. Tsokos, “A weighted moving average process for forecasting,” Journal of Modern Applied Statistical Methods, vol. Vol 16, No. 2, 2008. R. N. Calheiros, R. Ranjan, C. A. F. D. Rose, and R. Buyya, “Cloudsim: A novel framework for modeling and simulationof cloud computing infrastructures and services,” CoRR, vol. abs/0903.2525, 2009.

Chirag Rathod,IJRIT

284