task scheduling algorithm based on improved min-min algorithm in cloud computing environment

5
Task scheduling algorithm based on improved Min—Min algorithm in cloud computing environment Guan Wang 1,a , Haicun Yu 1,b 1 Modern education technology center, TangShan Vocational&Technical College, TangShan, China a [email protected], b [email protected] Keywords: Cloud Computing; Task scheduling; Min-Min algorithm; Algorithm simulating; Abstract. Task schedule algorithms directly related to the speed and quality of schedule. Min-Min algorithm always completes the shortest total completion time task first, and has the characteristic of simple and shortest completion time. This paper research scheduling algorithm based on Min—Min algorithm. The result shows that the proposed algorithm is efficient in the cloud computing environment. Introduction Cloud computing is on behalf of a new critical Point on network computing value. It provides a higher efficiency, great expandability and faster, easier software development. The center content is the new programming model, the new IT infrastructure as well as the realization of the new business model. For the past few years, the giant such as IBM, Google, Amazon, Microsoft have set foot in cloud computing, and provide a lot of based on cloud computing services. Resources management and operation scheduling on cloud computing is the key technology of cloud computing. At present cloud computing operation scheduling algorithm between design doesn't have a unified standard. If the simple distribution scheduling methods, such as common rotary method, the weighted rotary method, minimum load priority, weighted minimum load priority method, hash method, etc,are adopt, it is difficult to achieve physical server load balance, and then can cause service performance imbalance and other related problems. Therefore this paper proposes an improved algorithm for cloud computing task scheduling strategy in order to maximize the efficiency of cloud computing environment. The overview of cloud computing The programming model analysis of cloud computing. The cloud computing environment mostly used distributed MapReduce computing model. The task will be divided into several subtasks automatically to realize the task in large-scale computing node of the scheduling and distribution through the Map and Reduce two steps. Most of the information technology manufacturers proposed cloud plan used in programming model based on the idea of MapReduce development programming tools. MapReduce model is particularly applicable to production and processing large data set. MapReduce can be divided into two main stages: Map stages: firstly divide a big task into M pieces through the MapReduce library of the user program, the size of each piece general from 16 to 64 MB. Then parallelly execute at multiple worker, and output process after the intermediate file. Applied Mechanics and Materials Vols. 303-306 (2013) pp 2429-2432 Online available since 2013/Feb/13 at www.scientific.net © (2013) Trans Tech Publications, Switzerland doi:10.4028/www.scientific.net/AMM.303-306.2429 All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of TTP, www.ttp.net. (ID: 130.207.50.37, Georgia Tech Library, Atlanta, USA-13/11/14,08:51:50)

Upload: hai-cun

Post on 16-Mar-2017

215 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Task Scheduling Algorithm Based on Improved Min-Min Algorithm in Cloud Computing Environment

Task scheduling algorithm based on improved Min—Min algorithm in

cloud computing environment

Guan Wang 1,a , Haicun Yu 1,b

1Modern education technology center, TangShan Vocational&Technical College, TangShan, China [email protected], [email protected]

Keywords: Cloud Computing; Task scheduling; Min-Min algorithm; Algorithm simulating;

Abstract. Task schedule algorithms directly related to the speed and quality of schedule. Min-Min

algorithm always completes the shortest total completion time task first, and has the characteristic

of simple and shortest completion time. This paper research scheduling algorithm based on

Min—Min algorithm. The result shows that the proposed algorithm is efficient in the cloud

computing environment.

Introduction

Cloud computing is on behalf of a new critical Point on network computing value. It provides

a higher efficiency, great expandability and faster, easier software development. The center content

is the new programming model, the new IT infrastructure as well as the realization of the new

business model. For the past few years, the giant such as IBM, Google, Amazon, Microsoft have set

foot in cloud computing, and provide a lot of based on cloud computing services.

Resources management and operation scheduling on cloud computing is the key technology of

cloud computing. At present cloud computing operation scheduling algorithm between design

doesn't have a unified standard. If the simple distribution scheduling methods, such as common

rotary method, the weighted rotary method, minimum load priority, weighted minimum load

priority method, hash method, etc,are adopt, it is difficult to achieve physical server load balance,

and then can cause service performance imbalance and other related problems. Therefore this paper

proposes an improved algorithm for cloud computing task scheduling strategy in order to maximize

the efficiency of cloud computing environment.

The overview of cloud computing

The programming model analysis of cloud computing. The cloud computing environment

mostly used distributed MapReduce computing model. The task will be divided into several

subtasks automatically to realize the task in large-scale computing node of the scheduling and

distribution through the Map and Reduce two steps. Most of the information technology

manufacturers proposed cloud plan used in programming model based on the idea of MapReduce

development programming tools. MapReduce model is particularly applicable to production and

processing large data set.

MapReduce can be divided into two main stages:

Map stages: firstly divide a big task into M pieces through the MapReduce library of the user

program, the size of each piece general from 16 to 64 MB. Then parallelly execute at multiple

worker, and output process after the intermediate file.

Applied Mechanics and Materials Vols. 303-306 (2013) pp 2429-2432Online available since 2013/Feb/13 at www.scientific.net© (2013) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/AMM.303-306.2429

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of TTP,www.ttp.net. (ID: 130.207.50.37, Georgia Tech Library, Atlanta, USA-13/11/14,08:51:50)

Page 2: Task Scheduling Algorithm Based on Improved Min-Min Algorithm in Cloud Computing Environment

Reduce stages: summarily analysis processed results in Map stage, output the final treatment

results.

In MapReduce programming model, how to schedule many child tasks to achieve resource

dynamic load balancing, at the same time, also consider response time, and is still a complex

problem. Each task is subdivided into M pieces in the Map phase, and R pieces in the Reduce stage.

The number of M and R shall be greatly more than the number of worker machine; each worker

performs many different jobs.

The task scheduling in cloud computing environment. The background of Cloud computing

system has hundreds of thousands of ordinary or senior Server. the scheduler (or intermediate

server) in the center of task scheduling statistics attribute information of each resource to measure

calculation ability and communication ability of the resources nodes. the main parameters includes

CPU number m and processing power p (MIPS), in addition to the place the bandwidth of the

network B (decide the time from distribution to transfer to the node). At the same time it also

introduce a load balancing factor, its quantitative value using load effeciency of resources, is also

the ratio of the accomplished the mission of quantity and the distribution of the amount of the

task.The calculation formula is: Lr = La/Lf, thereinto, Lγ means load effeciency, La means the

quantity of the accomplished mission, Lf means the sum of the assigned calculation of the task.

The scheduler of dispatching center should constantly detect the resource load and the status of

the project, so that produce positive influence to the distribution of the next tasks. Whenever a new

task is completed, resources pheromone of higher task efficiency adds some, resources pheromone

lower task efficiency reduce some.

Task scheduling algorithm based on improved Min—Min algorithm in Cloud computing

environment

The analysis of the traditional Min - Min algorithm. Min - Min (Minimum - Minimum

completion time) algorithm belongs to heuristic dynamic task scheduling algorithm. It main goal is

to make a lot of tasks can assign to operate on the earliest resources. Min - Min algorithm is the one

of the basic task scheduling algorithm In the cloud computing environment.

In Min-min, the minimum completion time for each task is computed respect to all machine

resources. The task with the overall minimum completion time is selected and assigned to the

corresponding machine. The newly mapped task is removed to look for the smallest tasks for

distribution, and the process repeats until all scheduling task set is empty.

The limitations of the traditional Min - Min algorithm. According to scheduler’s prediction

finish time, traditional Min - Min algorithm mainly considerate to assign a lot of tasks to the fastest

machine, with the goal of completing the task in a minimum time and the fastest speed. In cloud

computing, facing special requirements of different task and the equal service requirements of the

different resources owner, just take an examination of the time factor without considering other

service quality requirement is obviously not enough. In general, the limitations of the traditional

Min - Min algorithm have two aspects, on the one hand, it is potential Load imbalance, resources

utilization rate is low; On the other hand, the traditional Min - Min algorithm doesn’t meet various

tasks of different service quality requirements.

The improved Min—Min algorithm based on QoS in Cloud Computing environment. In

line with the characteristics of the scheduling model in the cloud computing environment, this paper

puts forward a suitable scheduling algorithm for large-scale parallel processing system, namely,

makes full use of genetic algorithm model, fast search ability, at the same time uses the character of

Min—Min algorithm to establish a more optimal distribution scheduling strategy. The simulation

2430 Sensors, Measurement and Intelligent Materials

Page 3: Task Scheduling Algorithm Based on Improved Min-Min Algorithm in Cloud Computing Environment

experiments show that in task scheduling improved Min—Min algorithm can reduce the total

completion time tasks at the same time, balance the system load and improve the efficiency of the

task scheduling.

In the improved Min—Min algorithm all the publishing resources should provide its QoS

attribute description. The QoS attribute include scheduling completion time, reliability and price.

Scheduling completion time denoted by Finish Time, namely the time which the client user spend

from submitting task to finishing the task. The reliability is the ratio of the former n - 1 times call

success number and the total number of calls. The price is refers to the cost of the virtual machine

resources. Therefore, QoS can be described as a triad: Q = (Finish - Time, reliability, cost).

In the task scheduling process the QoS metric value and the load balancing factor need to be

considered.The scheduling task is divided into high QoS task queue and low QoS task queue. In the

execution scheduling, firstly, high QoS task is called by M in - M in scheduling algorithm.After

high QoS task completed, the low QoS task is scheduled. In the QoS task queues, according to the

load balance factor Virtual machine resources are sorted, it prefer light load of the virtual machine

resources for task allocation.

The simulation experiment based on CloudSim

CloudSim inherited GridSim programming model, support cloud computing modeling and

simulation. Its software structure frame and system structure components include four levels such as

SimJava, GridSim, CloudSim, UserCode. CloudSim provides resources monitoring and mapping

function from the host to the virtual machine. Comparing with the real cloud environment,

CloudSim can speed up the cloud computing algorithm test by systems modelling&simulation.

Fig. 1, The comparison of execution time Fig. 2, The comparison of load balance degree

In these experiment, the number of VM resources is two hundred, the number of the tasks is

under two hundred, and the initial reliability value random generate among 0.95-1.0.From the Fig. 1

and the Fig. 2, we found that the improved Min-Min algorithm has shorter completion time and is

able to balance the load very well compared to the Min-min algorithms. Among the Min - Min

algorithm each resource load is not balance, the main reason lies in that the shortest task is allocated

in the minimum load node, this leads to execution time long task is allocated to the larger load

nodes.

Applied Mechanics and Materials Vols. 303-306 2431

Page 4: Task Scheduling Algorithm Based on Improved Min-Min Algorithm in Cloud Computing Environment

Conclusions

According to the shortcomings of the traditional Min - Min algorithm,this paper proposes an

improved Min—Min algorithm based on QoS. The test results show that running performance is

improved greatly in the improved algorithm. About the task scheduling algorithm for Cloud

computing environment, how to optimize the influence of the data dependence to calculation

efficiency, how to meet higher scheduling efficiency, dynamic nature in cloud computing

environment, task QoS demand and so on, this will be the future scheduling algorithm a research

focus and development direction.

Acknowledgements

This research is funded by: the research program of Tangshan Municipal Science &

Technology Bureau: Ipv6 network application management system Based on cloud computing

platform (Grant No. 12140203A-1).

References

[1] Xian Jin,Yu Guangcheng. Research on scheduling algorithm based on cloud computing.

Computer&Digital Engineering.(2011).7:39-42.

[2] Yue Dongli,Liu Haitao,Sun Aobing. Research on IaaS public cloud computing platform

scheduling model. Computer Engineering and Design. (2011) Vol.32:1889-1897.

[3] Ma Jingyi,Sui Bing, Shu Wanneng. Task scheduling based on Min- Min genetic algorithm in

grid Computer Engineering and Applications. Computer Engineering and Applications.

(2008)44:102-104.

[4] The instruction manual of cloud computing simulation tool - CloudSim.

http://blog.csdn.net/rolinzcy12/article/details/4367631. (2009)

[5] Wang, S.C., Yan, K.Q., Liao, W.P., Wang, S.S. Towards a load balancing in a three-level cloud

computing network. Proceedings of 3rd IEEE International Conference on Computer Science

and Information Technology. (2010):108-113.

[6] Calheiros R N,Ranjan R,De Rose CAF,et al, loudsim:A novel Framework Ior Modeling and

Simulation of Cloud Computing Infrastructures and Services.GRIDS-TR-2009-1,Grid

Computing and Distributed Systems Laboratory. The University of Melbourne.Australia.March

13.(2009)

2432 Sensors, Measurement and Intelligent Materials

Page 5: Task Scheduling Algorithm Based on Improved Min-Min Algorithm in Cloud Computing Environment

Sensors, Measurement and Intelligent Materials 10.4028/www.scientific.net/AMM.303-306 Task Scheduling Algorithm Based on Improved Min-Min Algorithm in Cloud Computing

Environment 10.4028/www.scientific.net/AMM.303-306.2429