Task Scheduling using Hybrid Algorithm in Cloud Computing Environments

IOSR Journal of Computer Engineering (IOSR-JCE)
e-ISSN: 2278-0661,p-ISSN: 2278-8727, Volume 17, Issue 3, Ver. VI (May – Jun. 2015), PP 96-106
www.iosrjournals.org
DOI: 10.9790/0661-173696106 www.iosrjournals.org 96 | Page
Task Scheduling using Hybrid Algorithm in Cloud Computing
Environments
Ali Al-maamari1
, Fatma A. Omara2
1
(Department of Computer Science, Cairo University, Egypt)
2
(Professor, Department of Computer Science, Faculty of Computers & Information,
Cairo University, Egypt)
Abstract: The cloud computing is considered the latest network infrastructure that supports the
decentralization of computing. The main features of the Cloud are the possibilities for building applications
and providing various services to the end user by virtualization on the internet. One of the main challenges in
the field of the cloud computing is the task scheduling problem. Task scheduling problem concerns about the
dynamic distribution of the tasks over the Cloud resources to achieve the best results. Many of the algorithms
have been existed to resolve the task scheduling problem such as a Particle Swarm Optimization algorithm
(PSO). The PSO is a simple parallel algorithm that can be applied in different ways to resolve the task
scheduling problems. In this paper, a task scheduling algorithm has been proposed to the independent task over
the Cloud Computing. The proposed algorithm is considered an amalgamation of the PSO algorithm and the
Cuckoo search (CS) algorithm; called PSOCS. To evaluate the proposed algorithm, the cloudsim simulator has
been used. The experimental results show the reduction of the makespan and increase the utilization ratio of the
proposed PSOCS algorithm compared with PSO algorithms and Random Allocation (RA).
Keywords: Cloud Computing, PSO algorithm, CS algorithm, Task Scheduling, simulation
I. Introduction
Cloud Computing has become the fast spread in the field of computing and industry in the last few
years. As part of the services offered by the Cloud Computing is the new possibilities for building applications
and providing various services to the end user by virtualization on the internet.
On the other hand, many challenges need to be solved in the cloud computing. The two important
issues used in this paper are load balancing and Task scheduling. Load Balancing; is the process of distributing
the load statically or dynamically, among various nodes of a distributed system to improve both resource
utilization and job response time while avoiding a situation where some of the nodes are heavily loaded and
other nodes are idle or doing very little work [1]. According to task scheduling; an appropriate number of tasks
would be scheduled over the virtual machines. Task scheduling is the most important issue in the cloud
computing, because the user will have to pay for resource using on the basis of time. The goal of task
scheduling is to spread the load evenly between the system by maximizing utilization and reducing task
execution Time [2].
Moreover, there are many load balanced algorithms have been proposed in different environments.
However, with the cloud environment, some additional challenges have been existed and must be addressed. In
the Cloud Computing, efficient allocation of tasks, and the contract of such a request will be processed as
efficiently as possible[3].
The work in this paper, an algorithm based on particle swarm optimization algorithm and cuckoo
search algorithm has been presented to optimize the task scheduling problem in the Cloud environment to
minimize the completion time and increase the utilization ratio using open source Cloudsim3.0.3 simulator. The
cloudsim is a cloud computing simulation software developed by Gridbus project team and the grid Laboratory
of the University of Melbourne in Australia. The Cloudsim can run on Linux and Windows systems[4].
Kennedy and Eberhart[5] have presented a as a self-adaptive global search based optimization
technique. The algorithm is similar to other population based algorithms as Genetic algorithms (GA) without
direct re-combination of individuals of the population. Instead, it relies on the social behavior of the particles. In
every generation, each particle adjusts its trajectory based on its best position (local best) and the position of the
best particle (global best) of the entire population. These concepts increase the stochastic nature of the particle
and converge quickly to a global minimum with a reasonable good solution. Particle Swarm Optimization has
become popular because of its simplicity and its effectiveness in a wide range of application. Some of the
applications that have used PSO to solve NP-Hard problems like Scheduling problem[6], task allocation
problem[7], the data mining problem[8], environmental engineering problem[9]. On the other hand, Yang, X.-S.
And S. Deb.[10] Have introduced Cuckoo Search algorithm, which is considered a meta-heuristic optimization
algorithm based on the behavior of the cuckoo bird. There are some similarities between CS algorithm and hill-

Task Scheduling using Hybrid Algorithm in Cloud Computing Environments
climbing algorithms in respect with some large scale randomization. But, these two algorithms are in essence
very different. Firstly, CS algorithm is population-based algorithm in a way similar to GA algorithm and PSO
algorithm, but it uses some kind of selection similar to that used in harmony search. Secondly, the
randomization is more efficient as the step length is heavy-tailed, and any large step is possible. Finally, the
number of tuning parameters is less than in GA and PSO, and thus CS can be much easier adapted to a wider
class of optimization problems.
The proposed task scheduling is considered an amalgamation of tow algorithms PSO and CS.
The rest of this paper is organized as follows;Section 2 provides related work. Section 3 describes the
main components task scheduling. Sections 4 describe the intelligent PSO algorithm. Sections 5 describe the
intelligent CS algorithms, and sections 6 describe our proposed algorithm in detail. Section 7 presents the
comparative study between our proposed algorithms with PSO and RA algorithms. Finally presents the
conclusions and future research directions.
II. Related Works
Task scheduling has been a significant research topic whose objective is to ensure that every computing
resource is distributed efficiently and fairly and in the end improves resource utility. In traditional computing
environments of distributed computing, parallel computing and grid computing, a set of static and dynamic and
mixed scheduling strategies have proposed.
Suraj Pandey et al. [11]have proposed a particle swarm optimization (PSO) based heuristic to schedule
applications to cloud resources that optimizes computation cost and data transmission cost. The PSO based
mapping algorithm has much lower cost as compared to another algorithm called (Best Resource Selection)
based mapping. The PSO algorithm is used for a workflow application through the difference of its computation
and communication costs. The results show that PSO can achieve cost savings and good distribution of the
workload onto resources.
B. Radojevic[12] has considered to fix the load balancing algorithms and overcome the defects round
robin algorithm, Which is the very famous algorithm and it is working on the basis of a conversion session in
the application layer. The main feature of the algorithm is to improve communication time between the
customer and the node in the cloud computing. If the connection time has exceeded the threshold standard, then
the relation between the customer and the node will be finished, the task is to provide some other node by using
round robin rules. The approach sends a request to a node with less number of communications.
Lee[13], has proposed a dynamic load balancing algorithm on the basis of an existing algorithm called
WLS (weighted least connection).According to this algorithm is that assigning tasks to node according to the
number of connections that exist for this node. It is comparing between a set of connections from each node in
the cloud and the task assigned to the node with the lowest number of connections. Nonetheless, weighted least
connection would not take into account the capabilities of each node, such as processing speed and storage and
capacity and bandwidth.
Ren[14], An improved has been in reduced to the algorithm of the WLC [13], with taking into account
the time series and trials, called Exponential Smooth Forecast based on Weighted Least Connection
(ESWLC).The ESWLC steps are; 1) builds the conclusion of the assignment of a specific task to the node after
having a number of tasks assigned to that node and identify node capabilities. 2) Builds a decision on the basis
of experience of the node’s memory, CUP memory, number of connections and the amount of disk space
currently being used. 3) Then predicts which node is to be chosen on the basis on exponential smoothing.
Hai Zhong1 et al.[15]have proposed algorithm to present an optimum, scheduling algorithm to achieve
the optimization for cloud scheduling. In this algorithm an Improved Genetic Algorithm (IGA) is used for the
automatic schedule policy. It is used to increment the utilization ratio of resources and speed.
Xin Lu, ZilongGu.[16], Dong, Wang, D. [17] and Song, X., L. [18] are proposed a Load balancing task
scheduler balance the entire system load while trying to minimizing the makespan of a given task set. They are
used two different load balancing scheduling algorithms based on the solution of ant-colony optimization
(ACO) technique, which aims to minimize the completion time based on pheromone.
Pooranian et al.[19], have proposed a task-scheduling technique for grid computing based on a merge
PSO with the gravitational emulation local search (GELS) algorithm, which aims to minimize makespan and the
number of tasks that fail to meet their deadlines.
III. The Principle And Components Algorithm
Figure 1 shows an overview of the algorithm model, the model consists of three modules, the first
module is the scheduler system, which consists of a scheduler, and the second module is the application which
represents the set of the cloudlets and a set of the virtual machines. The third module is the Mapping Algorithms
(MA) which estimate the expected time for each cloudlet to allocate on each virtual machine, and it is assumed
that these values are available to the scheduler. The cloudlets exepected time have been stored in an m × n

matrix, where m is the number of virtual machines and n is the number of cloudlets. Obviously, n/m will
generally be greater than 1, with more cloudlets than virtual machines, so that some machines will need to be
assigned multiple cloudlets. The Estimated Running Time (ERT) is defined as the time the of executing task j on
resource r [20]. Each column i of the expected running times (ERT) matrix contains estimates of the expected
running time (ERT) of each cloudlet j on machine i.
Figure 1: The Component of PSOCS algorithm
The main target of allocating on a virtual machine is to reduce Makespan, Makespan can be defined as
the overall task completion time. We denote completion time of task Tio n VMja s CTij . Hence, the makespan
is defined as the following equation[21]:
Makespan = CTmax i, j | i ϵ T, i = 1,2, … n and jϵ VM, j = 1,2, … m 1
WhereCTmax i, j is the maximum which can be defined as the time for completing cloudlet i on a virtual
machinej.
3.1 Mathematical Modules
Let VM = VM1, VM2, … . VMm be the number of m virtual machines that must be process n tasks
represented by the group T = T1, T2, … Tn . Each virtual machines are parallel and independent, the schedule
independent tasks to theseVMs, the Processing of that task on a virtual machine cannot interrupt (i.e) Non-
preemption, we denote end time of a task Ti byCTij . Our aim is to reduce the Makespan which can be denoted
asCTmax , the run time of each task for each virtual machine must be calculated for the purpose of scheduling, If
the processing speed of virtual machine VMjis PSj, then the processing time for cloudlet Pican be calculated by
equation.(2)[22]:
Pij = Ci PSj (2)
Where Pij the processing is time of task Piby virtual machineVMj and Ciis the computational complexity of the
task Pi[22]. The values obtained from equation (2), are stored in the runtime matrix.
Th processing time of each task in the virtual machine can be calculated by equation (3)
Pj = Pij = 1, … , m (3)
n
i=1
By minimizingCTmax , Eqs. (1), (2) and (3),we can calculated that (4).
Pij ≤ CTmax (4)
n
i=1
By considering of the load balancing, the tasks will be transferred from one VM to other in order to reduce
CTmax ,as well as, response time. The processing time of a task varies from one VM to another based on the
virtual machines speed. In case of transferring, the completion time of a task may vary because of load
balancing, optimally.
CTmax = {maxi=1
n
CTi , maxj=1
n
Pij
n
i=1
} (5)
The main target our proposed algorithm is that tasks should be allocated virtual machine and
minimized makespan and maximizing the resource utilization. Our proposed tasks scheduling algorithm is

considered an extension of an existing PSO algorithm and the Cuckoo Search algorithm merged with the
concept of Cuckoo Search.
IV. The Particle Swarm Optimization Algorithm
In this section, the algorithm for task scheduling based on PSO will be explained.
According to the PSO system, the population is individuals of particles. Each particle forms a candidate
solution to problem space. Particles are initialized randomly with a fitness value. This value is calculated by a
fitness function to be best solution in each generation. Each particle knows best position pbest and the best
position so far among the complete set of particles gbest.The pbest of a particle is the best result that is
calculated by the fitness value, while gbest is the best particle in terms of fitness for all population. The
evaluation is implemented to number of repetitions[23, 24]. When iteration optimization process is bag an, each
of the iteration of the velocity and the positions of all particles are updated according to the equation. (6, 7)[23].
vi k + 1 = w ∗ vi k + c1 ∗ r1 ∗ pbesti k − xi k + c2 ∗ r2 ∗ gbesti k − xi k (6)
xi k + 1 = ki k + vi k + 1 (7)
Where,vi(k) andxi(k)are the velocity of particle𝐢 at iteration k, r1and r2are random numbers with a regular
distribution in the period between 0 and 1.c1 andc2 are learning factors called the cognition and the social
parameter and w is inertia weight can be dynamically different through applying an annealing scheme for the w
setting of the PSO. w is decreased and contributed to convergence. In general the inertia weight w is set
according to equation. (8)[25].
w = wmax −
wmax − wmin
itramax
× itra (8)
Wherewmax is the initial value, wmin is the final value of the weight coefficient, itramax is the largest possible
number of repetitions. The performance of the PSO algorithm is improved heavily by changing inertia using the
updated equation (6) and (7). The pseudo code of the PSO algorithm is presented in figure (2)[23].
Figure2: The PSO algorithm Pseudo code
V. Cuckoo Search Algorithm
The cuckoo is considered special bird because it has many of the characteristics that distinguish it from
other birds.
It is characterized by aggressive breeding strategy. Cuckoo lays their eggs in the nest of another
species, sometimes the cuckoo's egg in the host nest is discovered may lead to the removal of other eggs or
abandons the nest and builds their own brood somewhere else in [26].
5.1 Cuckoo Behavior
Some cuckoo species have evolved in such a way that female parasitic cuckoos are often very
specialized in the

mimicry incolor and pattern of the eggs of a few chosen host species. This reduces the probability of
eggs being abandoned and increases their reproductively.Parasitic cuckoos often choose a nest where the host
bird just laid its own eggs. In general, the cuckoo eggs hatch slightly earlier than their host eggs.Once the first
cuckoo chick is hatched, the first instinct action will be taken is to evict the host eggs by blindly propelling the
eggs out of the nest, which increases the cuckoo chick’s share of food provided by its host bird[26].Cuckoo
characteristics could be described, as a model for good behavior other animals have extensive use in computing
Intelligence Systems[27].
5.2 Cuckoo Rules & Parameters
To simplify the principle of the novel Cuckoo Search (CS) algorithm, three exemplary rules can be used [26].
1. Each cuckoo lays one egg at a time, and dumps it in a randomly chosen nest.
2. The best nests with high quality of eggs (solutions) will carry over to the next generations.
3. The number of available host nests is fixed, and a host can discover an alien egg with a probability pa ∈ [0,
1]. In this case, the host bird can either throw the egg away or abandon the nest so as to build a completely
new nest in a new location.
As a further approximation, this last assumption can be approximated by a fraction pa of the n nests
being replaced by new nests with new random solutions at new locations. For a maximization problem, the
quality or fitness of a solution can simply be proportional to thevalue of its objective function. Other forms of
fitness can be defined in a similar way ofthe fitness function in genetic algorithms and other
evolutionarycomputation algorithms. For simplicity, the following Simple protests is used that each egg in the
nest is a solution, the cuckoo egg represents the new solution, and the goal is to use a new and potentially better
solution (Cuckoo) to replace worse solutions that are in the nests. Of course, this algorithm can be extended to
more complex situation where each nest has multiple eggs represent a set of solutions. When generating new
solutions X(t+1)
for a cuckoo i, a Lévy flight is performed using the equation (10) [26].
X(t+1)
= X(t)
+ α ⊕ Lévy λ (9)
Where X(t)
represents the current location, α > 0 is the step size, which should be related to the scales
of the problem that the algorithm is trying to solve. In most cases, α = 1, and λ Є(0,3) are used. The equation(9)
is in core stochastic equation for a random walk which is similarly of a markov chain who’s next location
(status) depends on two parameters; current location (the first term in equation. 9) and the possibility of
transmission (the second term in Eq. 10).The product represents entry-wise multiplication[10]. Something it
is similar to entry-wise product as in PSO algorithm, but random walk through a Lévy flight is a random walk in
which the step-lengths are distributed according to a heavy-tailed probability distribution. After a large number
of steps, the distance from the origin of the random walk tends to be a stable distribution[28]. Here, the
consecutive jumps (steps) of a cuckoo essentially form a random walk process which obeys a power-law step
length distribution with a heavy tail. Taking into account basic three rules described above.
The first proposed algorithm is called the Cuckoo Search (CS).The Pseudo code for the CS algorithm is
shownin Figure 3 [10]. According to the CS algorithm, an initial set of nests, which represent the solutions, are
randomly generated. These solutions are then updated over multiple generations. The process of updating an
individual solution is as follows; a random nest is chosen, and a new solution is generated by random-walking
from this previous solution. This new solution can then replace a different randomly chosen solution if it has a
fitness value better than the original. After this possible replacement of a solution, all of the nests are ranked by
fitness and the worst fraction of the nests is replaced with random solutions. This combination of mechanisms
allows the solutions to search locally and globally at the same time for the optimal solution[10].

Figure2: Pseudo code for The CS algorithm
VI. The Proposed Task Scheduling ALGORITHM
There various search stages in the PSO algorithm. These stages are close to the optimum stage with
their pbestandgbest values. The main drawback of the PSO algorithm is its weakness of local searches because
there is a possibility of becoming trapping in a local search in the last repetition. The problem is not taken place
in the global search, because the PSO algorithm is always trying to reach for the solutions that have better
fitness functions in the search problem space. Therefore, the PSO algorithm is unable to recognize and avoid
local optima. As a result, the PSO algorithm may become trapped in local optima and have a reduction
convergence ratio in the late repetition process. Then, Cuckoo Search (CS) algorithm has been used to
overcome the local optima problem by amalgamating with the PSO algorithm.
According to the proposed task scheduling algorithm, the PSO algorithm has been used as the main search
algorithm, while the CS algorithm is used to improve the population. There are two reasons for using both
algorithms. First, it needs an algorithm based on a population to search the entire cloud space for this problem.
Second, the cloud environment is dynamic, so the scheduling algorithm must be fast enough to adapt with the
natural cloud environment and must be able to converge faster than other algorithms. Moreover, although the
PSO algorithm is weak for local searches, by combining the PSO algorithm with CS algorithm is considered
powerful in searches addresses this problem. Because CS actions have a high computational cost for each
particle in the PSO search and the cloud scheduler should execute quickly, CS is run on the global result of the
last iteration of the PSO. That is, an initial solution for CS which is provided by PSO during the mix search
process.
One of the most major challenges to apply CS and PSO algorithms in the task scheduling problem is
that how to enter a schedule as a search solution, find appropriate maps among problem solutions, and how the
CS nests with the PSO particles. In our proposed PSOCS algorithm, each particle represents a possible solution
for the task assigned using an array of n elements, where all elements randomly produce integer values between
1 and m. Figure 4 shows the assignment of ten tasks to five virtual machines. For example, in Particle 1 or nest
1, tasks T1, T5 and T3 are assigned to VM1 and tasks T2 , T6 and T4 are assigned to VM2 and T7 is assigned to
VM3 and T8 and T9 are assigned to VM4, and T10 is assigned to VM5.
Figure 4: Particle Representation
6.1 The PSOCS Task Scheduling Algorithm
The methodology of the PSOCS process is illustrated by the flowchart shown in Figure 5.The main flowchart
steps are:
1- Initialization by dividing the application into set of cloudlet.
2- Create cloudlet and virtaul machine.
3- Schedule the cloudlet using the proposed algorithm.

4- Assign the cloudlte T to the scheduler, using parallel computation to compute the cloudlets with the help of
the workers
5- Find out the cloudlet execution time Tet and Total timeTt.
6-Termination check when the entire cloudlet Tt has been assigned to the scheduler, the algorithm terminates
Step 3 is the main process of the algorithm.
Figure 5 presents the flowchart of the proposed PSOCS task scheduling algorithm.
Figure 5: Flowchart of the Proposed PSOCS Scheduling Algorithm
VII. Performance Evaluation
Cloudsim is cloud computing, simulation software developed by the Gridbus project team and the grid
Laboratory of the University of Melbourne in Australia. CloudSim has been used to implement the proposed
PSO-CS task scheduling algorithm. Also, a comparative study has been done to evaluate the performance of the
proposed PSO-CS algorithm with respect to the existed Random Allocation algorithm(RA) [29], and the original
PSO algorithm. This simulation mainly validates the advantage of the makespan and the resource utilization
among these scheduling algorithms in the Cloud Computing environment.
7.1. The Implementation Results
By using equation (1), the simulation results to evaluate the makespan of the three algorithms; Random
Allocation (RA), original PSO, and PSO-CS; using 5, and 10 Virtual machines, and 10, 20, 30, and 40 tasks are
described in Figures (4, 5) and Table (1,2) respectively. According to the simulation results, it is found that the
proposed approach (PSO-CS) outperforms the RA and the original PSO algorithms.
Table 1 represents the makespan of RA, PSO and PSO-CS task scheduling algorithms using 5 virtual
machines. Figure 4 illustrates the comparison of Makespan using the algorithms. The X-axis represents the
number of cloudlets (tasks) and Y-axis represents the execution time (Makespan) in seconds. According to the
results in figure 4, it is found that makespan of our proposed (PSOCS) is significant decreased by considering
more number of cloudlets. The proposed PSO-CS algorithm provides best results relative to RA and PSO
algorithms. By RA and PSO vs. PSO_CS improvement; 21.45% and 26.65% respectively.
Figure 4: Execution time of all Cloudlet when No. VMs (5)

Table1: Compared scheduling algorithm with Execution Time (Sec)
RA PSO PSO_CS VM Cloudlet
3 3.11069307 2.391129032
5
10
4.6 5.230736842 4.268631579 20
14.1 12.07002398 9.076738609 30
17.1 22.13103448 13.21939655 40
Table 2 represents the makespan of RA, PSO and PSO-CS task scheduling algorithms using 10 virtual
machines. By RA and PSO vs. PSO_CS improvement; 27.02% and 22.48% respectively.
Table2: Compared scheduling algorithm with Execution Time (Sec)
RA PSO PSOCS VM Cloudlet
2.2 2.21908549 1.48072562
10
10
5.9 4.48764805 3.57886497 20
7 7.15238095 6.09310345 30
9.9 9.70897704 7.61503132 40
Figures (4, 5) show comparisons of makespan when number of VMs is varied of 5 and 10 for RA, PSO
and PSOCS. In all the cases algorithms it is clearly from the graph that our algorithm performs better other
algorithm.
The simulation result utilization is described as in Tables (3, 4) and Figure. (6, 7), Tables (3, 4) show
comparison of utilization with RA, PSO and PSOCS when number of virtual machines is varying once when
No.VM is five other once when NO. VM is ten. Fig (6, 7) illustrates the comparison of utilization using
algorithms discussed previously. The X-axis represents the number of cloudlets and Y-axis represents the
utilization ratio. It is clearly from the graph that PSOCS is more efficient when compared with other algorithms.
We used around 40 tasks for our comparisons.
Figure 5: Execution time of all Cloudlet when No. VMs (10)
Figures (6, 7) show comparisons of utilization when number of VMs is varied of 5 and 10 for RA, PSO
and PSOCS. In all the cases algorithms it is clearly from the graph that our algorithm performs better other
algorithm.
Table3: Comparison three algorithms with Utilization vs. NO. VMs(5)
0.62 0.50629 0.82404
5
10
0.7521739 0.724139489 0.888377435 20
0.5858156 0.644754177 0.871109968 30
0.49239767 0.388064025 0.639837228 40
Table4: Comparison three algorithms with Utilization vs. NO. VMs(10)
0.368182 0.38077399 0.58296606 10
0.355932 0.41487024 0.49601614 20
0.457143 0.36903696 0.49878417 10 30
0.39596 0.3348205 0.45243068 40

Figure 6: Calculate Utilization of all Cloudlet when No.VMs (5)
Figure 7: Calculate Utilization of all Cloudlet when No.VMs (10)
In the experiment, the number of the virtual machine is five and the number of the Cloudlets is varying.
And show the experimental results with clarification of the process of distribution of tasks on virtual machine in
Table 5 and Table 6. In the table shows the first column is the number of cloudlet. The second column is the
number of virtual machines. The third column is the cloudlets running time, the fourth column is the start time,
and the fifth column is the end time.
From table 5 It can be seen that because of scheduling PSO four tasks are Assigned to the virtual
machine No. 4, is not assigned to any task to the virtual machine No. 1.These lead to load imbalance and the
total execution time of ten cloudlets is the longest.
Table5: Distributed cloudlet on VM in PSO algorithm
PSO
Cloudlet NO VM NO Run Time Start Time Finish Time
7 4 0.688118812 0.1 0.788118812
2 4 0.706534654 0.788118812 1.494653466
5 4 0.732079208 1.494653466 2.226732674
1 2 0.784835165 0.1 0.884835165
0 4 0.883960396 2.226732674 3.11069307
4 2 0.910549451 0.884835165 1.795384616
3 2 0.962637363 1.795384606 2.758021969
9 3 0.990554415 0.1 1.090554415
6 0 1.007526882 0.1 1.107526882
8 0 1.309408602 1.107526882 2.416935484
In Table 6, it can be seen that because of the PSOCS, each virtual machine has cloudlets to run. The load is
very balanced. The execution time of the ten cloudlet is less, In Table 5 maxtime is 3.11069307either table 6
maxtime is 2.391129032.
Table5: Distributed cloudlet on VM in PSOCS algorithm
PSOCS
Cloudlet NO VM NO Run Time Start Time Finish Time
7 4 0.688118812 0.1 0.788118812
1 4 0.707128713 0.788118812 1.495247525
6 3 0.769609856 0.1 0.869609856
2 2 0.784175824 0.1 0.884175824

5 2 0.812527473 0.884175824 1.696703297
0 4 0.883960396 1.495247525 2.379207921
8 3 1.000205339 0.869609856 1.869815195
4 0 1.113709677 0.1 1.213709677
3 0 1.177419355 1.213709677 2.391129032
9 1 1.296774194 0.1 1.396774194
Figure (a, b) shows the comparison of makespan and utilization to number of virtual machines when
number of VMs are 5 vs. a set varied of cloudlets. Also in Figure 8 (c, d) illustrates comparison of makespan
and utilization when number VMs are 10. Results illustrate that PSOCS is more performance when compared
with RA and PSO algorithms.
Figure 9 shows the makespan with different numbers of virtual machine (i.e., VM). As we can see,
performance is also affected by the number of virtual machines .In specifically, the more virtual machine
available, the easier to find an optimal solution and reduce makespan.
Figure 8: (a, c) Comparison between Makespan of numbers of Cloudlets vs. number of
the VMs.
(b, d) Comparison between Utilization numbers Cloudlets vs. Number of the VMs.
Figure 10 shows the utilization with different numbers of virtual machine (i.e., VM). As we can see,
utilization is also affected by the number of virtual machines. In specifically, the more virtual machine available,
the utilization ratio increases.

VIII. Conclusion And Future Work
The work in this paper present a combine task scheduling algorithm based on the PSO and CS. This
algorithm is called PSOCS. The target of the combine is to minimize makespan and increase utilization ratio of
the application workflows in the Cloud computing. Where increases utilization ratio is obtained through
decreases the completion time of the executed tasks on a virtual machine. The core principle of the proposed
PSOCS algorithm is based on a combine of PSO and CS algorithms. The combine PSO with CS algorithms
performs better for local searches. Because the CS used in the local search instead other local search algorithms
such as hill-climbing. The combine algorithm finds better solutions than other algorithms. Generally, a
comparison of the performance of PSO-CS with existed RA and the original PSO algorithms through a
simulation experiment using cloudsim shows that PSO-CS performs better than that RA and the original PSO
algorithms. In future work, we plan to improve PSO and CS algorithms in order minimum execute time, and
maximum of utilization resource.
References
[1]. Mishra, R. and A. Jaiswal, Ant colony optimization: A solution of load balancing in cloud. International Journal of Web &
Semantic Technology (IJWesT), 2012. 3(2): p. 33-50.
[2]. Uma, J., V. Ramasamy, and A. Kaleeswaran, Load Balancing Algorithms in Cloud Computing Environment-A Methodical
Comparison. International Journal of Advanced Research in Computer Engineering & Technology (IJARCET) Volume. 3.
[3]. Nishant, K., et al. Load balancing of nodes in cloud using ant colony optimization. in Computer Modelling and Simulation
(UKSim), 2012 UKSim 14th International Conference on. 2012. IEEE.
[4]. Calheiros, R.N., et al., CloudSim: a toolkit for modeling and simulation of cloud computing environments and evaluation of
resource provisioning algorithms. Software: Practice and Experience, 2011. 41(1): p. 23-50.
[5]. Kennedy, J. and R. Eberhart. Particle swarm optimization. in Neural Networks, 1995. Proceedings., IEEE International Conference
on. 1995.
[6]. Yu, B., X. Yuan, and J. Wang, Short-term hydro-thermal scheduling using particle swarm optimization method. Energy Conversion
and Management, 2007. 48(7): p. 1902-1908.
[7]. Yin, P.-Y., et al., A hybrid particle swarm optimization algorithm for optimal task assignment in distributed systems. Computer
Standards & Interfaces, 2006. 28(4): p. 441-450.
[8]. Sousa, T., A. Silva, and A. Neves, Particle swarm based data mining algorithms for classification tasks. Parallel Computing, 2004.
30(5): p. 767-783.
[9]. Lu, W., et al., Analysis of pollutant levels in central Hong Kong applying neural network method with particle swarm optimization.
Environmental Monitoring and Assessment, 2002. 79(3): p. 217-230.
[10]. Yang, X.-S. and S. Deb, Engineering optimisation by cuckoo search. International Journal of Mathematical Modelling and
Numerical Optimisation, 2010. 1(4): p. 330-343.
[11]. Suraj, P., et al., A Particle Swarm Optimization (PSO)-based Heuristic for Scheduling Workflow Applications in Cloud Computing
Environments. 2010.
[12]. Radojevic, B. and M. Zagar. Analysis of issues with load balancing algorithms in hosted (cloud) environments. in MIPRO, 2011
Proceedings of the 34th International Convention. 2011. IEEE.
[13]. Lee, R. and B. Jeng. Load-balancing tactics in cloud. in Cyber-Enabled Distributed Computing and Knowledge Discovery
(CyberC), 2011 International Conference on. 2011. IEEE.
[14]. Ren, X., R. Lin, and H. Zou. A dynamic load balancing strategy for cloud computing platform based on exponential smoothing
forecast. in Cloud Computing and Intelligence Systems (CCIS), 2011 IEEE International Conference on. 2011. IEEE.
[15]. Zhong, H., K. Tao, and X. Zhang. An approach to optimized resource scheduling algorithm for open-source cloud systems. in
ChinaGrid Conference (ChinaGrid), 2010 Fifth Annual. 2010. IEEE.
[16]. Lu, X. and Z. Gu. A load-adapative cloud resource scheduling model based on ant colony algorithm. in Cloud Computing and
Intelligence Systems (CCIS), 2011 IEEE International Conference on. 2011. IEEE.
[17]. Li, K., et al. Cloud task scheduling based on load balancing ant colony optimization. in Chinagrid Conference (ChinaGrid), 2011
Sixth Annual. 2011. IEEE.
[18]. Song, X., L. Gao, and J. Wang. Job scheduling based on ant colony optimization in cloud computing. in Computer Science and
Service System (CSSS), 2011 International Conference on. 2011. IEEE.
[19]. Pooranian, Z., et al., An efficient meta-heuristic algorithm for grid computing. Journal of Combinatorial Optimization, 2013: p. 1-
22.
[20]. Blythe, J., et al. Task scheduling strategies for workflow-based applications in grids. in Cluster Computing and the Grid, 2005.
CCGrid 2005. IEEE International Symposium on. 2005. IEEE.
[21]. Brucker, P. and P. Brucker, Scheduling algorithms. Vol. 3. 2007: Springer.
[22]. Kruekaew, B. and W. Kimpan. Virtual Machine Scheduling Management on Cloud Computing Using Artificial Bee Colony. in
Proceedings of the International MultiConference of Engineers and Computer Scientists. 2014.
[23]. Visalakshi, P. and S. Sivanandam, Dynamic task scheduling with load balancing using hybrid particle swarm optimization. Int. J.
Open Problems Compt. Math, 2009. 2(3): p. 475-488.
[24]. Selvarani, S. and G.S. Sadhasivam. Improved cost-based algorithm for task scheduling in cloud computing. in Computational
intelligence and computing research (iccic), 2010 ieee international conference on. 2010. IEEE.
[25]. Uma, S., et al., A hybrid PSO with dynamic inertia weight and GA approach for discovering classification rule in data mining.
International Journal of Computer Applications, 2012. 40(17).
[26]. Yang, X.-S. and S. Deb. Cuckoo search via Lévy flights. in Nature & Biologically Inspired Computing, 2009. NaBIC 2009. World
Congress on. 2009. IEEE.
[27]. Rambharose, T. and A. Nikov, Computational intelligence-based personalization of interactive web systems. 2011.
[28]. Brown, C.T., L.S. Liebovitch, and R. Glendon, Lévy flights in Dobe Ju/’hoansi foraging patterns. Human Ecology, 2007. 35(1): p.
129-138.
[29]. Xu, X., et al., Cloud task and virtual machine allocation strategy in cloud computing environment, in Network Computing and
Information Security. 2012, Springer. p. 113-120.

Task Scheduling using Hybrid Algorithm in Cloud Computing Environments

Recommended

More Related Content

What's hot (18)

Viewers also liked (17)

Similar to Task Scheduling using Hybrid Algorithm in Cloud Computing Environments (20)

More from iosrjce (20)

Recently uploaded (20)

Task Scheduling using Hybrid Algorithm in Cloud Computing Environments