IRJET- A Particle Swarm Optimization Algorithm for Total Cost Minimization in the Cloud Manufacturing

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 06 | June 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 2617
A Particle Swarm Optimization Algorithm for Total Cost Minimization
in the Cloud Manufacturing
S.A.A. Jude1, S. Selva Prabhu2, P.Thangababu3, S. Muruga Perumal4, M. Sathish5
1Assistant Professor, Mechanical Engineering, PSN College of Engineering and Technology, Tirunelveli, India
2Research Scholar, PSN College of Engineering and Technology, Tirunelveli, India
3,4,5UG Student, BE Aeronautical Engineering, PSN College of Engineering and Technology, Tirunelveli
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Cloud manufacturing is a hybrid model that
provides both hardware and software resources through
computer networks. Data services (hardware) together with
their functionalities (software) are hosted on web servers
rather than on single computers connected by networks.
Through a computer, a browser and an internet connection,
each user accesses a cloud platform and asks for specific
services. This model creates a brand new opportunity for
enterprises. It is largely impossible to carry out some
manufacturing tasks without the support of suppliers,
contractors and business partners. There is a need to connect
manufacturers together to sharerisks, benefits, competiveness
and costly resources. Morethanjustdeployingmanufacturing-
related software applications in the computing Cloud, Cloud
Manufacturing is an integrated solution that provides a pool
of machine capabilities provided by Cloud participants.
Key Words: Cloud manufacturing, Particle swarm
optimization algorithm, Cloud computing
1. INTRODUCTION
Cloud Manufacturing is a model for enabling ubiquitous,
convenient and on-demand network access to a shared pool
of configurablemanufacturingresources (e.g.,manufacturing
software tools, manufacturing equipment, and
manufacturing capabilities) that can be rapidly provisioned
and released with minimal management effort or service
provider interactions. Like Cloud Computing concept,
manufacturing infrastructure, platform and software
application in Cloud Manufacturing can be offered as a
service to a Cloud User. By extending the concept to a
broader scope, all the production objectsandfeaturescanbe
treated as a service, hence everything as a service. The rest
of this section discusses the Cloud Manufacturing structure
and related technologies.
Cloud concept presents a promising future for computing
business and the same can be said for manufacturing
business. Cloud Manufacturing is described as a computing
and service-oriented manufacturing model developed from
existing advanced manufacturing models (e.g., Application
service provider, Agile Manufacturing, Networked
Manufacturing, and Manufacturing Grid), enterprise
information technologies under the support of cloud
computing, Internet of things, virtualization and service-
oriented technologies, and advanced computing
technologies. In this paper, we focus on minimizing the total
execution cost of applications on these resources provided
by Cloud service providers, such as Amazon and GoGrid3.
We achieve this by using a meta- heuristics method called
Particle Swarm Optimization (PSO).
2. PROBLEM FORMULATION
The mapping of tasks of an application workflow to
distributed resources can have several objectives. We focus
on minimizing the total cost of computation of an application
workflow. Figure 5 depicts a workflow structure with five
tasks, which are represented as nodes. The dependencies
between tasks are represented as arrows.
Figure1 An example workflow, compute nodes (PC) &
storage (S).
This workflow is similar in structure to our version of the
EvolutionaryMulti-objectiveOptimization(EMO)application.
The root task may have an input file (e.g. f.in) and the last
task produces the output file (e.g. f.out). Each task generates
output data after it has completed (f12, f13, ..., f45). These
data are used by the task’s children, if any. The numeric
values for these data is the edge-weight (ek1,k2) between
two tasks k1 Є T and k2 Є T . The figure also depicts three
compute resources (PC1, PC2, PC3) interconnected with
varying bandwidth and having its own storage unit (S1, S2,
S3). The goal is to assign the workflow tasks to the compute
resources such that the total cost of computation is
minimized.
Task 1 Task 2 Task 3 Task 4 Task 5
PC2 PC2 PC3 PC3 PC1

A sample particle for the workflow
Task 1 Task 2 Task 3 Task 4 Task 5
PC
1 3 3 2 1
2 3 1 1 3
1 2 3 3 2
3 2 3 1 2
3 2 2 3 1
Table 1: The Taskflow
3. PARTICLE SWARM OPTIMIZATION ALGORITHM
PSO was first introduced by Kennedy and Eberhart as an
optimization method for non-linear functions with
continuous variables. The initial intention of PSO was to
simulate the social behavior of flocking birds searching for
food by means of exchanging knowledge among flock
members. By applying simple formulae, Kennedy and
Eberhart developed an optimization algorithm that mimics
this knowledge sharing. Each individual in the flock was
represented by a point in a two- dimensional space, and
future movement of each point in the search space is
determined using a combination of previous experience of
the individual, and of other individuals in the group.
The PSO provides a population-based search procedure in
which the individuals, calledparticles,change theirpositions
with time. Each particle adjusts its position according to its
own best experience and the best experience of neighboring
particles.
The PSO heuristic is applied by first generating a number
of random solutions (or positions of particles) in thesolution
space. PSO searches for optimal solution by updating
generations. Each particle is updated by means of two ‘best’
values, namely pbest (p_k
i) and gbest (p_k
g) in successive
iteration. The pbest (p_k
i) is the best solution, a particle has
achieved so far. The gbest (p_k
g) is the best solution obtained
so far by any particle in the population. The quality of each
particle position is then evaluated based on the objective
function. To proceed from iteration k to the next iteration
k+1, the velocity of a particle i is calculated using the
following equation (1).
The new position of the particle is obtained by equation (2),
Where s_k
i is particle i position in the current iteration k. ω
can be expressed by the inertia weights approach, Ca and Cb
are the acceleration constants which influence the
convergence speed of each particle, and rand( ) is a
random number between 0 and 1. In equation (1), the first
part represents the inertia of the previous velocity, the
second part is the “cognition” part which represents the
private thinking by itself, and the third part is the “social”
part which represents the cooperation among the particles.
If the summation in equation (1) causes the velocity vki, on
that dimension, to exceed Vmax, then vki will be limited to
Vmax. Vmax determines the resolution with which regions
between the present position and the target position are
searched. If Vmax is too large, the particlesmightflyoverthe
past good solutions. If Vmax is too small, the particles may
not explore sufficiently beyond local solutions. Usually the
range of particle is taken as the Vmax. In this project,Vmax=
6. The constants Ca and Cb represent the weighting of the
stochastic acceleration terms that pull each particle toward
p_ki and p_kg positions. Low values allowparticlestoroamfar
from the target regions before being tugged back. On the
other hand, high values result in abrupt movement toward
or passed the target regions. Hence the acceleration
constants Ca and Cb are often set to be 2.0 according to the
past experiences. Suitable selection of inertia weight ω
provides a balance between global and local explorations,
thus requiring less iteration on average to find a sufficiently
optimal solution. As originally developed, ω often decreases
linearly from about 0.9 to 0.4 during a run. In general, the
inertia weight ω is set according to the following equation:
ω Inertia weight;
v_ki Velocity of particle i in the current iteration k;
v_(k+1)i Velocity of particle i in the next iteration k+1;
p_ki The best solution that particle i reached throughout
iterations 1,2, …….k,;
p_kg The best solution that the group has reached
throughout iterations 1,2, …..k,;
s_ki Particle i position in the current iteration k;
s_(k+1)iParticle i position in the next iteration k+1;
rand() Uniformly distributed random number generated
between 1 and 3;
Ca& Cb Learning factors.
Where itermax represents the maximum number of
iteration, and iter is the current number of iterations.
itermax depends on the problem to be optimized. Here, it is

taken as 500. Moreover, ωmax and ωmin are the maximum
and minimum weight values respectively.
4. Proposed PSO Algorithm
Set particle dimension as equal to the size of ready
tasks
Initialize particles position randomly For each
particle, calculate its fitness valueIf the fitnessvalueisbetter
than the previous best pbest, set the current fitness value as
the new pbest.
After Steps 3 and 4 for all particles, select the best
particle as gbest.
For all particles, calculate velocity using Equation 1
and update their positions using Equation 2.
If the stopping criteria or maximum iteration is not
satisfied, repeat from Step3.
5. Experimental Evaluation
In this section, we present the metric of comparison, the
experiment setup and the results.
Five possible combinations of tasks are considered in the
initialization module. Matrix [1] shows theinitial population
consists of five particles. Each and everyparticleintheinitial
population has a single row.
[1]
A. Evaluation module
Matrix [1] is called s_ki and the manufacturing system
efficiency is calculated for all particles. The maximum total
indexing time in seconds is corresponding to the second
particle. Hence, the second particle is selected and is called
gbest. pkg for the first iteration. The p_kg is then converted
into the same size of ski (i.e., 5 x 3), by repeating the same
row.
B.Scheduling heuristic Algorithm.
Calculate average computation costofall tasksinall
compute resources. Calculate average cost of
(communication/ size of data) between resources Set task
node weight wkj as average computation cost Set edge
weight ek1,k2 as size of file transferred between tasks
Compute PSO({ti}) /*a set of all tasks i ∈ k*/repeat for all
“ready” tasks {ti} ∈ T do Assign tasks {ti} to resources {pj}
according to the solution provided by PSO end for Dispatch
all the mapped tasks Wait for polling time Update the ready
task list Update the average cost of communication between
resources according to the current network load Compute
PSO({ti}) until there are unscheduled tasks
C. Scheduling Heuristic
We calculate the average computationcostall tasksonall the
computer resources. This cost can be calculated for any
application by executing each task of an application on a
series of known resources. It is represented as TP matrix in
Table 2. As the computation cost is inversely proportional to
the computation time, the cost is higher for those resources
that complete the task quicker. Similarly, we store the
average value of communication costbetween resources per
unit data, represented by PP matrix in Table 2, described
later in the paper. The cost of communication is inversely
proportional to the time taken. We also assumeweknowthe
size of input and output data of each task.In addition, we
consider this cost is for the transfer per second (unlike
Amazon CloudFront which does not specify time for
transferring). The initial step is to compute the mapping of
all tasks in the workflow, irrespective of their dependencies
(Compute PSO(ti)).
This mapping optimizes the overall cost of computing the
workflow application.Tovalidatethedependencies between
the tasks, the algorithm assigns the “ready” tasks to
resources according to the mappinggivenbyPSO.By“ready”
tasks, we mean those tasks whose parents have completed
execution and have provided the files necessary for the
tasks’ execution. After dispatching the tasks to resources for
execution, the scheduler waits for polling time. This time is
for acquiring the status of tasks, which is middleware
dependent
TP(5x3)=
PC1 PC2 PC3
T1 1.23 1.12 1.15
T2 1.17 1.17 1.28
T3 1.13 1.11 1.11
T4 1.26 1.12 1.14
T5 1.19 1.14 1.22
Table 2 The Cost of execution of Ti at PCj
Depending on the number of tasks completed, the ready list
is updated, which will now contain the tasks whose parents
have completed execution. We then update the average
values for communication between resources according to
the current network load.Asthecommunicationcostswould
have changed, we recompute the PSO mappings. Also, when
remote resource management systems arenotabletoassign
task to resources according to our mappingsduetoresource
unavailability, the recomputation ofPSOmakestheheuristic

dynamically balances other tasks’ mappings (online
scheduling).
PP (3x3)=
PC1 PC2 PC3
PC1 0 0.17 0.21
PC2 0.17 0 0.22
PC3 0.21 0.22 0
Table 3 The Cost of Comunication between PCi & PCj
Based on the recomputed PSO mappings, we assign the
ready tasks to the compute resources. These steps are
repeated until all the tasks in the workflow are scheduled.
6. CONCLUSIONS
In this work, we presented a scheduling heuristic based on
Particle Swarm Optimization (PSO).We used the heuristic to
minimize the total costofexecutionofapplicationworkflows
on Cloud computing environments. Weobtainedtotal costof
execution by varying the communication cost between
resources and the execution cost of compute resources. We
compared the results obtained by our heuristic against“Best
Resource Selection” (BRS) heuristic.
We found that PSO basedtask-resource mappingcanachieve
at least three times cost savings as compared to BRS based
mapping for our application workflow. In addition, PSO
balances the loadoncomputeresources bydistributingtasks
to available resources. The heuristic we proposed is generic
as it can be used for any number of tasks and resources by
simply increasing the dimension of the particles and the
number of resources, respectively.
REFERENCES
[1]XunXun “ From cloud computing to cloudmanufacturing”
Department of Technical Engineering, University of
Auckland, Auckland1142, NewZealandRobotics and
Computer-Integrated Manufacturing 28 (2012) 75–86
[2]OmidFatahiValilai,MahmoudHoushmand“Acollaborative
and integrated platform to support distributed
manufacturing system using a service-oriented approach
based on cloud computing paradigm “Robotics and
Computer-Integrated Manufacturing 29 (2013) 110–127
[3]Alexander Verla, Armin Lechlera, Stefan Wesnerb,
Andreas Kirstädterc, JanSchlechtendahla, Lutz Schubertd,
Sebastian Meierc “An approach for a cloud-based machine
tool control” Procedia CIRP 7 (2013) 527 – 532
[4]RaminVatankhah “Online velocity optimizationofrobotic
swarm flocking using particle swarm optimization (PSO)
method” IntJAdvManufTechnol March 24-26 2009,40 (9)
:1257-1284
[5]SandhyaraniBiswas, Mahapatra S.S, “An improved
metaheuristic approach for solving the machine loading
problem in flexible manufacturing systems”, International
Journal of Services and Operations Management, Volume 5,
Number 1, 2009, pp. 76 – 93
[6]HameshbabuNanvalaPonnambalam S.G, Low SengKiat,
“SolvingMachineLoadingProbleminFlexibleManufacturing
Systems Using Particle Swarm Optimization”, World
Academy of Science, Engineering and Technology 39 , 2008,
pp.14-19.
[7]Mahapatra et al., proposed, a metaheuristic approach
based on PSO to solve the machine loading
problem.2008International Journal of Services and
Operations Management
[8]Xi VincentWang and XunW.Xu et al., proposed an
interoperable solution for Cloud manufacturing.
[9]G.H. Liu, Y.S. Wong, Y.F. Zhang and H.T. Loh proposed
Modeling cloud data for prototype manufacturing.
Department of Mechanical and Production Engineering,
Journal of Materials Processing Technology, Vol.138(2003),
No. 1-3, p. 53-57
[10]L. Grandinetti , O. Pisacane and M. Sheikhalishahi
proposed an approximate constraint method for a multi-
objective job scheduling in the cloud Future Generation
Computer SystemsVolume 29, Issue 8, October 2013, Pages
1901–1908.

IRJET- A Particle Swarm Optimization Algorithm for Total Cost Minimization in the Cloud Manufacturing

More Related Content

What's hot (18)

Similar to IRJET- A Particle Swarm Optimization Algorithm for Total Cost Minimization in the Cloud Manufacturing (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- A Particle Swarm Optimization Algorithm for Total Cost Minimization in the Cloud Manufacturing