Dynamic flow sheduling based on particle swarm optimization in virtual machine.
Cloud computing is a computing concept that private public and hybrid cloud connected to the large systems. It can be delivers to the mutual computer processing resources on demand process. It is universal, on-demand access to a share the pool of computing properties are promptly provisioned and unrestricted with the minimal management effort. Cloud computing can have the smart features of virtualization technology, it is consolidation, separation, passage and overhang or continue support. In model of computing, the features are scalability are provided to the new type of building blocks called virtual machines (VMs). VMs can have the own scheduling tasks and profits. Cloud computing relies on sharing the property that can achieve consistency. The central supporting knowledge for the cloud computing is virtualization. Virtualization software separates a somatic computing device into single or more 'virtual' devices that are easily used to achieve good perform computing tasks. Virtualization essentially scalable system of operating system creating a multiple free computing devices, shiftless computing resources can be allotted used more efficiently.
Static scheduling is the tasks to mainframes to done before program execution start. In that sequence of concerning task and the execution times of the processing resources is assumed as compile time. It is a static scheduling methods are processor non primitive. The main advantage of the scheduling methods of static is to overhead of the scheduling process is incur at accumulate time, efficient execution time more background compared to the dynamic scheduling methods but it can be only at the preliminary stage. Dynamic scheduling is containing the arrangement of processes during on the performance time. It is redistribution method to performed by transmitting tasks from the a unlimited deal loaded processors to the lightly loaded processors are called load balancing. That the balancing is want to refining the presentation of the submission. That the load balancing processes may perhaps the centralized in one or more scattered between the processing elements that can be participated in to the capacity balancing development. More joined policies could also subsist. For model, the information policy can be centralized however the transfer and placement policies could be distributed. In that place whole processors transmit their contents information to the dominant processor and accept system load information from that processor. The flexibility intrinsic in dynamic load balancing allows for adjustment to the unpredicted application requirements at run-time.
Dynamic load balancing is predominantly useful in a system consisting of a network of workstations the most important performance goal is maximizing consumption of the giving out power instead of minimizing execution time of the applications.  DDFS is to schedule the flow to improve overall link utilization although achieving closely related load values among all aggregate switches. DDFS has the potential to produce better link utilization inherent policies are information, transfer, placement policy are used to allocated how they are identify and transferred it can be used in policy.
The PSO algorithm works a populace means swarm and candidate solutions is called particles. The particles are moved about in the search-space due to simple methods. The particles movement is directed own the best position in the search-space besides the entire swarm's best identified position. It can be improved positions that are discovered by the controller the schedules the swarm . It can be repeated and doing it is expected, but not definite, that a acceptable solution will be finally be exposed. In particle swarm optimization that present a particle swarm optimization (PSO) based scheduling heuristic algorithm for data concentrated applications that takes account in to both division cost and data transmission cost . The cost savings is used the PSO that is compared to using active Best Resource Selection (BRS) algorithm. That our results defined that can be achieve considerably more than three times total investments as associated to BRS good distribution of workload on to resources, that using PSO based scheduling heuristic.joint static and Dynamic traffic scheduling in data centre network are traffic model and scheduler structure both called Generalized Load Balanced Scheduler (GLOBS) it can be improve the performance.data center interconnection design problem can be occurred.
The assignment scheduling problem can't be solving in polynomial point [5,7..]. So this problem can be used by the heuristics algorithms particle swarm optimization that can be minimize the execution time of particle swarm optimization, that the parallel version of the algorithm is mainly used in the optimization the parallel version of the algorithm is used to the parallel particle swarm optimization that can be reduces single of the finishing time of responsibilities in cloud location. [6, 12 ...]In PSO the algorithm is opted to minimize costs in calculation to minimizing execution time.
In scheduling methods that can be order to minimize the cost of the processing that formulate a model for task scheduling and propose a particle swarm optimization (PSO) algorithm it is based on small position value rule.
A Hyper-Heuristic scheduling Algorithm for cloudthis work that to advance to the performance rule based scheduling algorithm used to the cloud computing system by using simple easy to implement. Hyper heuristic scheduling algorithm is identify better scheduling solution for the cloud computing. The algorithm can be mostly used in the variety recognition and enlargement finding operators. That hyper heuristic scheduling algorithm  used of the scheduling.  Feedback-Based Scheduling for Load-Balanced that virtual ouput queuing that only a single data buffer can be required that better delay throughput.
OpenFlow based scheduling schemes transmit the data's are statically at initial stage of the data transmission . That system performance is poor during the dynamically flow distribution changing the network states in data center. The scheduling approaches of big data center in cloud means the datacenter contains data broker and number of virtual machines and number of cloudlet also. In that process number of files are allocated to the virtual machines dynamically. Allocation process are under the Dynamic Load Balanced Scheduling (DLBS), in this approach for maximizing performance of virtual machine.Dynamically load balancing scheduliding that dater center create and broker created then this jobs are loaded in to virtual machine. loud let files that secheduling and allocated in job scheduling methods. virtual machine that allocated normally so manys datas or jobs are waiting and time delay problem occur. so that Partical Swarm Optimization(PSO) algorthim that define solving the problems in load balancing in virtual machine. pso goal is to work can be minimize the jobs execution time.
Network Model in network model a Cloud Data Center network is an objectiveless graph. G (V, E) Graph V is union of the switch set (vs.) E defined union of Es' Open flow Based Data Centre Networks. In this scheduling method, the types of network can be defined. Three-layer non-blocking Fully Populated Network (FPN) Three-layer Fat-Tree Network (FTN)
Fully Populated Network:
The fully populated network is widely used in data center network.switches are connected by each level. Fully populated network has a good connectivity process.
The fat tree network (FTN) network model is for large-scale system-level network. Fat-Tree Network has good extendibility.
B. Dynamical load-balanced scheduling algorithms:
Estimate Schemes Performance of mid-way load balanced scheduling measured by the network throughput the realistic condition of two models are Improved One-Hop DLBS-FPN 2) Multi-Hop DLBS-FTN. In dynamic scheduling methods are initialization,bandwidth monitoring, rescheduling process are mainly used in scheduling.
Flows are initiated and distributed symmetrically among all hosts. Transmits packet with equal probability.
Flow generation is distributed in the intra-pod that connect with directly. The improvement in transmission delay under uniform and semi-uniform patterns is lower than that in center-based pattern.
The center based pattern most unbalanced traffic pattern 80% of data flows are generated by single host.
In dynamically load balancing scheduling methods the data can be dynamically balanced in virtual machine. But in virtual machine data that more comes execution time is taken more that process it to optimization techiest is used by the way of particle swarm optimization is developed.
D. Particle Swarm Optimization:
The particle swarm optimization is swarm based technics that defined in to separate of both types. In virtual machine that particle that workload can be calculated. That the fitness value can be individually updated. Then the swarm process find the minimum fitness value and it can allocated the files to minimum fitness value based on the capacity.
The list of modules used in the work are listed below and explained in detail as follows
> Data center and broker creation
> Virtual machine creation
> PSO Algorithm
A. Data Center and Broker Creation:
The data center creation is huge collection of networked computer servers normally used by organization used for the isolated loading processing. Datacenter can create for virtual more data. The data center that can be contain number of data center name, datacenter id, host id, bandwidth, RAM. A broker is an individual person that arranges transactions between a request and a response from the data center. it has been created for users to communication. The broker data center and the broker id that can handle the data dynamically in database.
B. Virtual Machine Creation:
Virtual machine creation that number of jobs that can be allocated for the process. Virtual machine runs inside of the computer that sharing the list with other VMs. it process files. That VM has an owner that can submit files to the VM to be executed. Number of files that can be stored in cloudlet.selected files are stored into the virtual machine. Files that are allocate to virtual machine dynamically in this process any files are allocate any one of the virtual machine that find the performance of virtual machine.
C. PSO Algorithm:
Particle Swarm Optimization (PSO) algorithm is a swarm-intelligence-based to estimate on nondeterministic. The PSO algorithm maintains several possible solutions at one time through both iteration of the algorithm that each result can be evaluated by an objective function to determine its qualification of each solution is defined the particle in the fitness land that means search space. The particles "wing" or "group" through the search space to catch the maximum value returned by the objective function. The algorithm have three global variables are goal value or state global best value indicating the particle's data is currently closest to the target Stopping value indicating the algorithm when should stop and the target isn't found. The particle consist of defines three important data defines a possible results. And the velocity value indicated the data how much changed. A personal best value representing the closest the particle's data ever come to the board. The PSO algorithm is used to find the fitness values for the process to avoid unwanted data access It also used to increases the optimization methods to save time. PSO is prepared for a for random particles that means solution it can searched for optimization. in that iteration that particle may have the two best values. The first values is fitness value that p best (personal best) and the anther value of the tracked PSO is defined by gbest (global best). The particle takes the topological best value is local best (lbest). Finding the best values the particle informs that velocity and position of equation 1 and 2.
v = v + c1 * rand() * (pbest - present) + c2 * rand() * (gbest - present) (1) present = present + v (2) > v defines velocity > present denoted current particle solution > rand() defines random values among (0,1) > c1, c2 defines learning factors > pbest, gbest are the stated before For every particle Modify particle END Do For each particle Calculate capability value If the fitness value is better than the best capability value (pBest) in antiquity set current value as the new pBest End Choose the particle oh the best fitness value of particles as the gBest For both particle Estimate particle velocity according comparison (1) Inform particle position according comparison (2) End
Particle velocity can be defined in the maximum that can be analysis Vmax. in that sum of rushing cause the velocity on that dimension to go above Vmax. The parameter can identified from users. in that dimension is limited in the velocity of Vmax.
In this dynamical flow scheduling that time usage of the process time. Data that access on the on delay time for data transmission of the jobs allocated timing and the process time and delay timing is defines balance remaining time usage defined in the fig. 2
The fig defines the traffic balancing in virtual machine by the fitness values of the particle swarm allocated the jobs.
According the scheduling methods there are many scheduling approach and methods used in the data center creation. But some of the scheduling methods are not used in the flow scheduling that performance is low of them. Open Flow scheduling methods statically stage that data can be only at starting stage of data transmission. Dynamically load balancing flow scheduling approaches. The data can be maximizing the network throughput loaded in balancing workload dynamically way is using data center creation through virtual machine. The datacenter contains data through virtual machine using Dynamically Load balanced scheduling approaches. In this dynamically load balancing flow scheduling approach we get good performance of the virtual machine. particle swarm optimization algorithmic be solve the problem in load balancing while virtual machine access and the PSO algorithm works to minimize the execution time of the jobs in virtual machine.
[1.] Feilong Tang Member, Laurence T. Yang Senior Member, Can Tang, Jie Li Senior Member and MinyiGuo Senior Member, 2016. A Dynamical and Load-Balanced Flow Scheduling Approach for Big Data Centers in Clouds IEEE Transaction On Cloud Computing pp: 1-14.
[2.] Elina Pacini, Cristianmates, Carlos Gracia Garino, 2014. Dynamic Scheduling based on Particle Swarm Optimization for Cloud-based Scientific Experiments, CLEI ELECTRONIC JOURNAL, 14(1): 1-12.
[3.] Awad, A. I., N. A. El-Hefnawy, H. M. Abdel_kader, 2015. Enhanced Particle Swarm Optimization For Task Scheduling In Cloud Computing Environments ELSEVIER International Conference on Communication, Management and Information Technology (ICCMIT) pp: 920-929.
[4.] Suraj Pandey1, LinlinWu1, Siddeswara Guru2, Rajkumar, 2014. A Particle Swarm Optimization (PSO)-based Heuristic for Scheduling Workflow Applications in Cloud Computing Environments
[5.] Vidhya1, M., N. Sadhasivam, 2015. Parallel Particle Swarm Optimization for Task Scheduling in Cloud Computing International Journal of Innovative Research in Science Engineering and Technology, 4: 6.
[6.] Hadi Salimi, Mahsa Najafzadeh and Mohsen Sharifi, 2012. Advantages, Challenges and Optimizations of Virtual Machine Scheduling in Cloud Computing Environments International Journal of Computer Theory and Engineering, 4: 2.
[7.] Lizheng, Guo, 2012. Task Scheduling Optimization in Cloud Computing Based on Heuristic Algorithm JOURNAL OF NETWORKS, 7: 3.
[8.] Lizheng Guo, Shuguang Zhao, Shigen Shen, Changyuan Jiang, 2012. Task Scheduling Optimization in Cloud Computing Based on Heuristic Algorithm", JOURNAL OF NETWORKS, 7: 3.
[9.] Bing Hu, Kwan L. Yeung, 2011. Feedback-Based Scheduling for Load-Balanced Two-Stage Switches.' IEEE Transaction on Networking, 18: 1077-1096.
[10.] Chun-weiTsai, Wei-chang Huang, 2014. A Hyper-Heruristic scheduling Algorithm for cloud'IEEE Transaction on Cloud Computing, 05: 1-14.
[11.] Jin, H., P. Deng, J. Liu et al. 2013. OpenFlow-Based Flow-Level Bandwidth Provisioning for CICQ Switches. IEEE Transactions on Computers, 62(9): 1799-1812.
[12.] Kotkondawar, R.R., 2014. A Study of Effective Load Balancing Approaches in Cloud Computing International Journal of Computer Applications (0975 - 8887) pp: 87: 8.
[13.] Al-Fares, M., S. Radhakrishnan, B. Raghavan, N. Huangand A. Vahdat, 2010. Hedera: Dynamic flow scheduling for data center networks. Proc. of Networked Systems Design and Implementation (NSDI) Symposium
[14.] Cao, Z., M. Kodialam and T. V. Lakshman, 2014. Joint Static and Dynamic Traffic Scheduling in Data Center Networks', in Proceedings of IEEE INFOCOM pp: 2445-2553.
[15.] Sourabh Bharti, K. K. Pattanaik, 2013. Dynamic Distributed Flow Scheduling with Load Balancin for Data Center Networks, Elsevier,
(1) M. Nisha and (2) P. Prem Kumar
(1) PG Student, Department of Computer Science and Engineering, K.L.N. College of Engineering, Affiliated to Anna University, Chennai.
(2) Professor, Department of Computer Science and Engineering, K.L.N. College of Engineering, Affiliated to Anna University, Chennai.
Received 28 January 2017; Accepted 22 March 2017; Available online 4 April 2017
Address For Correspondence: M.Nisha, PG Student, Department of Computer Science and Engineering, KLNCE, Pottapalayam, Sivagangai--630 612, Tamil Nadu, India.
Caption: Fig. 1: End-To-End Delay of Flow Scheduling
Caption: Fig. 2: Time usage
Caption: Fig. 3: Fitness of synthetic traffic
Caption: Fig. 4: response time of files
|Printer friendly Cite/link Email Feedback|
|Author:||Nisha, M.; Kumar, P. Prem|
|Publication:||Advances in Natural and Applied Sciences|
|Date:||Apr 1, 2017|
|Previous Article:||IOT based movement detection technique using pattern matching algorithm.|
|Next Article:||Design of IoT based secure water supply application using zone based clustering.|