Dynamic Load Balancing Techniques For Improving .
International Journal of Computer Applications (0975 – 8887)Volume 138 – No.3, March 2016Dynamic Load Balancing Techniques for ImprovingPerformance in Cloud ComputingSrushti PatelPG Student,S.P.College of engineering,Visnagar, 384315, IndiaHiren Patel, PhDProfessor, S. P. College ofEngineeringVisnagar,384315,IndiaABSTRACTCloud Computing is an emerging area in IT sector whichenables a wide range of users to access distributed, scalable,virtualized hardware and/or software, applications andplatforms are provided over the Internet. Cloud Computing isa shared pool of Configurable computing resources whichrequire the proper distribution of dynamic workload amongmultiple computers to ensure no single node is underloadedor overloaded. Load Balancing aims to reduce response timeof jobs, increase overall performance, reduce communicationcost of servers, Resource optimization, maintain cost ofVMs, Maximize throughput and avoid overload of any singlenode. In this paper we discuss the various techniques relatedto Load Balancing in Cloud Environment and further wepropose a modified agent based technique which is used forBalancing a load of the all host and also manage the newarrival jobs to increase the overall performance of system.KeywordsCloud Computing, Load Balancing techniques, Dynamicworkload Distribution, Resource utilization.1. INTRODUCTIONThe National institute of standard and technology(NIST) thatdefines the, "Cloud computing is a model for enablingconvenient, on-demand network access to a shared pool ofconfigurable computing resources (e.g., networks, servers,storage, applications, and services) that can be rapidlyprovisioned and released with minimal management effort orservice provider interaction. This cloud model is composedof five essential characteristics(On-demand self-service,Broad network access, Resource pooling, Rapid elasticity,Measured Service), three service models(SAAS, PAAS,IAAS), and four deployment models(Public, Private, Hybrid,Community) [1].This technology is used for spreading large amount ofdatasets and files around the world. Handling such type oflarge amount of datasets that require the some techniques foroptimizing the overall performance and user satisfaction[2].Therefore the load balancing in Cloud Computing is majorissues. Load balancing technique is used for improving theoverall performance of the system by Distributing the totalworkload among multiple computing resources or datacenters, network links, central processing units, disk drives,on the cloud Server, Processor. This load balancingtechniques helps to achieve optimal resource utilization,maximize throughput, minimize response time, and avoidoverload of resources.One of the crucial issue related with Load balancing is toremoving the condition in which some of the nodes are overutilized or some of the nodes are under utilized. ImproperNimisha PatelResearch Scholar,RaiUniversity, AhmedabadAssociate ProfessorS. P. College of EngineeringVisnagar,384315,Indiaworkload is distributed among the all computing resourcesand simultaneously removing a condition in which some ofthe nodes are over loaded while some others are underloaded. Other issues are like (a) resource utilization (b)scalability (c) energy efficiency (d) latency (e) throughput (f)performance (g) money (h) Achieving green computing.However with proper load balancing technique we canreduce resource consumption which is not only helps inreducing cost but making enterprises greener[3,10,11].The rest of this paper is organized as follows. Section 2presents the Background terminology with energy model.Section 3 presents the Related work or existing LoadBalancing Techniques. Section 4 presents the comparisonand discussion of load balancing techniques with the allperformance metrics. Section 5 proposed method andSection 6 presents the Conclusion and future work relatedwith this research.2. BACKGROUNDIn this Section we classify the Load balancing techniques inmainly two categories: static algorithms and dynamicalgorithms [6, 12] that have been developed for cloudcomputing which is shown in fig.1.Static load balancing algorithms assign the tasks to the nodesto process new requests. The process is based solely on priorknowledge of the node‟s properties and capabilities. Staticalgorithms do not change the attribute like node‟s processingpower, memory and storage capacity at run time[2].Load Balancing 1] Classification of Load balancing techniques.[13]Static load balancing algorithms assign the tasks to the nodesto process new requests. The process is based solely on priorknowledge of the node‟s properties and capabilities. Staticalgorithms do not change the attribute like node‟s processingpower, memory and storage capacity at run time[2]. Thereare some problem with static load balancing i.e. in the longrun, static weight cannot be corrected and the node is bound1
International Journal of Computer Applications (0975 – 8887)Volume 138 – No.3, March 2016to deviate from the actual load condition, resulting in loadimbalance so it can‟t handle long-connectivity applicationswell. Most of the Dynamic load balancing algorithms relieson a combination of knowledge and run-time properties.These algorithms assign the tasks to a node and maydynamically reassign them to another node based on theattributes gathered and calculated. Dynamic load balancingalgorithms are more accurate and could result in moreefficient load balancing. Such Algorithms require constantmonitoring of the nodes and task[6,12].Client 1Client 2Jitender grover et al.[6] proposed an Agent based dynamicload balancing(ABDLB) scheme for cloud computing. Aftercomparing with traditional algorithm the advantages ofABDLB algorithm is CPU time consumption is 1 unitwhereas in traditional algorithm CPU time consumption is 10unit.Client 3Client Job managerAnt colony optimizationGenetic AlgorithmsAgent Based DynamicLoad Balancing TechniquesPRSThrottledDeMSVirtual machine managerVM 1mechanism of natural selection strategy. The Advantages ofthis techniques is that it can handle a vast search space,applicable to complex objective function and can avoid beingtrapping into local optimal solution. It also guarantees theQOS requirement of customer job.VM 2VM 3Fig.2 Load Balancing techniques in cloud computing [8]Here fig.2 shows that the all Load balancing Techniqueswhich is used for balancing the overall workload of all VMsinto the system. Job manager having a several VMs, usingthis list of VM it assign the desire job to the appropriate VM.If not any VM is free at that time the job manager wait forthe client request and place that job into queue for the fastprocessing[8].A. Energy ModelEnergy is the capacity to do the work. Energy consumptionin data centers by computing nodes are mostly determinedby the physical resources such as CPU, memory, diskstorage, and network interfaces. This energy model is createon the basis of that processor utilization has a linearcorrelation with energy consumption. Measure the energyconsumption for a particular task, to use the information asits processing time and processor utilization is sufficient[15].For host Hi at any given time, the utilization Ui is Definedas,𝑀𝑖𝑈𝑖 Huangke Chen et al.[7] present a novel scheduling algorithmProactive and Reactive Scheduling(PRS) which isdynamically exploits for scheduling real time, aperiodic,independent task. Benefits associate with this algorithm is, Itcan prohibit propagation of uncertainties throughout theschedule, This design allows each task waiting on LQ to startas soon as its preceding task has finished, so the possibleexecution delay for anew task is removed. This designenables overlapping of communications and computationsoverlapped to save time and improve schedulingperformance[6,13], It also can reduce the overheads of tasktransfer among hosts when corresponding VMs need tomigrate.Vibhore Tyagi et al.[8] can propose the load balancingtechnique with throttled Algorithm to reduce the cost andresponse time across VM‟s in multi data center and optimizeresponse time service broker policy. Throttled policy definesthe work to finding the applicable virtual machine forassigning individual job.Yu Liu et al.[9] propose a DeMS consist of hybrid scheme oftask scheduling and load balancing technique having threealgorithms,(1)On-Demand Scheduling, (2)Querying andMigrating task(QMT), (3)Staged task migration(STM).4. COMPARISON AND DISCUSSION:Load balancing in cloud computing is distributing theworkload among the all node and transfer the load fromheavily node to idle node. It helps to improve response time,migration time, throughput, Resource utilization, scalabilityand overall performance.Table 1. Comparison Of Different Load BalancingTechniques With Various Metrics.Metrics/tech. Nature𝑢𝑖𝑗Environment𝑗 1ResponseTimeMigrationtimeResour PerformanceceutilizationLBACO[4]Dynamic decentralizedicLesslesshighHigh3. RELATED WORKThis section gives a brief review about the various existingLoad Balancing Algorithms Which is used for Balancing theoverall load of any Host in cloud computing environment.Kum li et al.[4] suggested an algorithm called LoadBalancing Ant Colony Optimization(LBACO) based on taskscheduling policy. This algorithm used to finding out theoptimal resource allocation in Dynamic cloud system incomplex network. The LBACO algorithm chooses optimalresources to perform tasks according to resources status andthe size of given task in the Cloud environment.Kousik Dasgupta et al.[5] proposed an algorithm known asgenetic algorithm based load balancing algorithm has beenused as a soft computing approach, which uses theABDLB[6]PRS[7]Dynam2
International Journal of Computer Applications (0975 – 8887)Volume 138 – No.3, March izedDynamicLesslessLesshighighhHighLessResponse time: It is the time interval between sendingrequest and receiving response. This time should beminimum for increasing the performance.Migration time: It is time taken to migrating the task ortransfer the task from one node to another. Minimum timehaving the maximum performance.Throughput: This metrics is used to estimate the task,whose execution complete successfully. For increasing theperformance, increase the value of this metrics.Resource utilization: it is used to insure that the utilizationof system resources. Better load balancing techniques givethe better resource utilization. Scalability: It determines the ability of the system toaccomplish load balancing algorithm with a restrictednumber of nodes.In First walk it moves from first server to lastserver and gathers information from all servers, formaking decision for Load Balancing and In Second walk it balances the host‟s load on thebasis of Standard deviation methodPerformance: It represents the effectiveness of the systemafter performing the load balancing algorithm. When theabove metrics satisfy optimally then the overall systemperformance will be increases.5. OUR CONTRIBUTIONGrover et al [6] proposed a system Architecture whichconsists „n‟ numbers of clients connected with cloud serviceproviders via internet and service provider consists VM,managed unit and „m‟ numbers of shared pool of resourceswhich are considering as servers. At the shared pool ofHosts, agent complete one cycle in two walks:Fig. 3 System Architecture[6]Here, our contribution has been illustrated by green boxes.We divide this architecture in three layers, User layer,Scheduling layer and Resource layer. Also we use the masterslave mechanism for migrating the jobs from overloadedserver to underloaded server and for managing a new arrivaljobs. Agent store the information of all host into that Slaveand at the end all the Slave Metadata is stored into theMaster.Agent Walk1:In the Agent walk1 Grover et al[6] describes the model interms of flow chart. The working of the model can beexplained in five steps.Step1: Agent is activated at any random server and findsnumber of jobs in queue at that server.Step2: Agent will repeat this process for all servers of sharedpool.Step3: After that it will calculate AVERAGE.Step4: On the basis of AVERAGE, it will sense the server‟sstatus in terms of overloaded and underloaded.Step5: Server‟s status will be decided as follows. AVERAGE, then transfer the server‟s status asoverloaded. If the number of jobs at ith server is less than theAVERAGE, then transfer the server‟s status asunderloaded.At the step1 we recommends to, Calculate,meanofutilizationusing3
International Journal of Computer Applications (0975 – 8887)Volume 138 – No.3, March 2016𝑀𝑖𝑆 𝑗 1Calculate,meanof𝐸 1𝑃putilizationofHi𝑋𝑘𝑘 1Whereas, 𝑋𝑘 is sum of utilization of VMs on hostHi in time frame k.P is the total time frames.Calculate, variance of utilization for host Hi1𝑉 𝑃 𝑢𝑖𝑗p𝑋𝑘 E2𝑘 1The standard deviation equals to the square root ofV.At the Step3 we contribute, instead of calculating theAVERAGE, we calculate the predicted utilization usingstandard deviation method[15]. Because of calculatingAVERAGE of all job into the queue it‟s not enough forfinding the server is overloaded or underloaded we use thisstandard deviation method. Understanding of my best whichis better than calculating only AVERAGE of queuing jobs.PU E S*StdDevAt Step4 we use „PU‟ for deciding the host is overloaded orunderloaded. If the If „PU‟ is greater than the currentutilization of host than host status is overloaded otherwiseunedrloaded and this information is stored into the slave.At the last step5 Information of all Slave is stored into theMaster.Agent Walk2:In the Agent walk2 Grover et al[6] describes the model interms of flow chart. The working of this model can beexplained below.Agent will start backtracking from last server to first serverfor balancing load of servers. At each server it will check thecondition. If the status of server is overloaded than transferthe jobs to underloaded server otherwise receive the jobsfrom overloaded server. Continue this process until the firstserver.Here, we recommends to use master slave mechanism.Master have the all information about the slave. Agent iscurrently at the Master and ready for balancing the load andalso assigning the new jobs. First, agent will check the stateof slave in master having the id and state. For each hostcheck the condition, if the host slave state overloaded thenmaster send a migration request to the slave. If slave send apositive response to the master then migrate jobs tounderloaded host. If the host slave state underloaded thenReceive jobs from “overloaded” host or new job is directlyassign on that host. Agent will perform this operation until itreaches at the first host with balancing all host‟s loadincluding first server also.6. CONCLUSION AND FUTUREWORKLoad balancing is the major issue in cloud environment.Cloud load balancing is the process of distributing workloadsacross multiple computing resources or data centers, networklinks, central processing units, disk drives, on the cloud. Inthis paper, we analyze various techniques for load balancingin cloud computing. We discussed the advantages anddisadvantage of this algorithm. Using this technique,improper workload can distributed among the all nodeswhich are idle. It help to achieve the user satisfaction byimproving the metrics like, Response time, migration time,throughput, resource utilization, Scalability, and overallperformance of the system. Also using this load balancingtechniques we can reduce the energy consumption andcarbon emission to making environment greener. Also wepropose a modified agent based technique which is used forBalancing a load of the all host and also manage the newarrival jobs.As a future research direction, we implement this techniquein real cloud environment and also we can design moreefficient algorithm which will maintain a better trade-offbetween all the metrics of algorithm using,a)A combination of two or more existing techniquesb)Improvement in one of the available techniques orc)Completely a new approach.7. REFERENCES[1] Mell, P., & Grance, T. (2011). The NIST definition ofcloud computing.[2] Nuaimi, K. A., Mohamed, N., Nuaimi, M. A., & AlJaroodi, J. (2012, December). A survey of loadbalancing in cloud computing: challenges andalgorithms. In Network Cloud Computing andApplications (NCCA), 2012 Second Symposium on (pp.137-142). IEEE.[3] Sreenivas, V., Prathap, M., & Kemal, M. (2014,February). Load balancing techniques: Major challengein Cloud Computing-a systematic review. In Electronicsand Communication Systems (ICECS), 2014International Conference on (pp. 1-6). IEEE.[4] Li, K., Xu, G., Zhao, G., Dong, Y., & Wang, D. (2011,August). Cloud task scheduling based on load balancingant colony optimization. In Chinagrid Conference(ChinaGrid), 2011 Sixth Annual (pp. 3-9). IEEE.[5] Dasgupta, K., Mandal, B., Dutta, P., Mandal, J. K., &Dam, S. (2013). A genetic algorithm (ga) based loadbalancing strategy for cloud computing. ProcediaTechnology, 10, 340-347.[6] Grover, J., & Katiyar, S. (2013, August). Agent baseddynamic load balancing in Cloud Computing. In HumanComputer Interactions (ICHCI), 2013 InternationalConference on (pp. 1-6). IEEE.[7] Chen, H., Zhu, X., Guo, H., Zhu, J., Qin, X., & Wu, J.(2015). Towards energy-efficient scheduling for realtime tasks under uncertain cloud computingenvironment. Journal of Systems and Software, 99, 2035.[8] Alakeel, A. M. (2010). A guide to dynamic loadbalancing in distributed computer systems. InternationalJournal of Computer Science and Network Security(IJCSNS), 10(6), 153-160.[9] Mata-Toledo, R., & Gupta, P. (2010). Green datacenter: how green can we perform. Journal ofTechnology Research, Academic and BusinessResearch Institute, 2(1), 1-8.4
International Journal of Computer Applications (0975 – 8887)Volume 138 – No.3, March 2016[10] Lee, R., & Jeng, B. (2011, October). Load-balancingtactics in cloud. In Cyber-Enabled DistributedComputing and Knowledge Discovery (CyberC), 2011International Conference on (pp. 447-454). IEEE.[11] Hu, M., & Veeravalli, B. (2013). Requirement-awarestrategies for scheduling real-time divisible loads onclusters. Journal of Parallel and Distributed Computing,73(8), 1083-1091.IJCATM : www.ijcaonline.org[12] Sinha, P. K. (1998). Distributed operating systems:concepts and design. PHI Learning Pvt. Ltd.[13] Cao, Z., & Dong, S. (2012, December). Dynamic VMconsolidation for energy-aware and SLA violationreduction in cloud Computing. In Parallel andDistributed Computing, Applications and Technologies(PDCAT), 2012 13th International Conference on (pp.363-369). IEEE.5
Dynamic Load Balancing Techniques for Improving Performance in Cloud Computing Srushti Patel PG Student, S.P.College of engineering, Visnagar, 384315, India Hiren Patel, PhD Professor, S. P. College of Engineering Visnagar,384315,India Nimisha Patel Research Scholar,Rai University, A
8. Load Balancing Lync Note: It's highly recommended that you have a working Lync environment first before implementing the load balancer. Load Balancing Methods Supported Microsoft Lync supports two types of load balancing solutions: Domain Name System (DNS) load balancing and Hardware Load Balancing (HLB). DNS Load Balancing
Bruksanvisning för bilstereo . Bruksanvisning for bilstereo . Instrukcja obsługi samochodowego odtwarzacza stereo . Operating Instructions for Car Stereo . 610-104 . SV . Bruksanvisning i original
III. LOAD BALANCING Load balancing techniques are used to make sure that each machine in the cloud datacenter performs approximately the equal number of tasks at any point of time [6]. The dynamic workload of the cloud is distributed evenly among the nodes available for processing. For this, load balancing is done. However, the load
Load Balancing can also be of centralized load balancing and distributed load balancing. Centralized load balancing typically requires a head node that is responsible for handling the load distribution. As the no of processors increases, the head node quickly becomes a bottleneck, causing signi cant performance degradation. To solve this problem,
Internal Load Balancing IP: 10.10.10.10, Port: 80 Web Tier Internal Tier Internal Load Balancing IP: 10.20.1.1, Port: 80 asia-east-1a User in Singapore Database Tier Database Tier Database Tier External Load Balancing Global: HTTP(S) LB, SSL Proxy Regional: Network TCP/UDP LB Internal Load Balancing ILB Use Case 2: Multi-tier apps
It is used for Balancing the load according to controller and according to flow of Data as well. Data Plane handle Link Load Balancing and Server Load Balancing. The Distributed multiple control architecture is subcategorized into Flat Architecture and hierarchical Architecture. It helps to explore new dimensions of load balancing. Figure 4.
load balancing degree and the total time till a balanced state is reached. Existing load balancing methods usually ignore the VM migration time overhead. In contrast to sequential migration-based load balancing, this paper proposes using a network-topology aware parallel migration to speed up the load balancing process in a data center.
Figure 1: Load Balancing Model based on [4]. 2.2 Load Balancing As cloud computing continues to grow, load balancing is essential to ensure that the quality of service isn't compro-mised for end users [4]. Load balancing is the process of distributing workload amongst a collection of servers in a data center.
