Partitioning Mobile Application By Cloud Computing Using A Linear Programming Algorithm In The Graph
| Autor | Mostafa Ahmadi Meshkani - Mohammad Hadi Yousefi - Mohammad Javad Rashidi |
| Cargo | Islamic Azad University - University, Kashan, Iran - University, Kashan, Iran |
| Páginas | 624-640 |
QUID 2017, pp. 624-640, Special Issue N°1- ISS N: 1692-343X, Medellín-Colombia
PARTITIONING MOBILE APPLICATION BY CLOUD COMPUTING USING A LIN EAR
PROGRAMMING ALGORITHM IN THE GRAPH
(Recibido el
15-06-2017.
Aprobado el
04-09 2017)
Mostafa Ahmadi Meshkani
Department of Computer
engineering, kasha n Branch,
Islamic Azad University,
Kashan, Ir an1
Ahmadi6468@gmail.com*1
Mohammad Hadi Yousefi
Department of Mechatronics,
kashan Bra nch, Islamic Azad
University, Kashan, Ira n2
Mhu320@yahoo.com2
Mohammad Javad Rashidi
Department of Mechatronics,
kashan Bra nch, Islamic Azad
University, Kashan, Ira n3
mjavadrashidi@yahoo.com3
Resumen: En el campo de la computación en nube, la falta de la misma capacidad para los dispositivos,
así como la falta de conocimiento claro sobre el número de dispositivos utilizados conducir a no utilizar
algoritmos prácticos con facilidad, por lo que el uso de algoritmo óptimo es adecuado cuando cualquier
dispositivo no puede ir Más allá de su capacidad máxima. También, en este trabajo, se ha considerado la
estabilidad de la partición. De hec ho, un subgrafo debe ser seleccionado en términos de conectividad de
modo que si un número de enlaces están desconectados, su estabilidad no se perderá y el programa se
ejecutará correctamente. En este trabajo, se presentó un método para dividir el gráfico de tareas de una
aplicación. Dado que el pr oblema de la partición gráfica a gran escala es de la dureza NP, el algoritmo
genético fue propuesto como una estructura selectiva en el método propues to. En el algoritmo genético,
tres criterios, incluyendo el costo, el tiempo de respuesta y la energía se utilizaron como un objetivo
combinado. El uso de la programación lineal influye efectivamente en el rendimiento óptimo del
algoritmo genético. Los res ultados del método propuesto muestran una reducción precisa del consumo de
energía y un tiempo de respuesta del 0,5% y del 3%, respectivamente.
Palabras clave: Algoritmo genético, programación lineal, particionamiento de gráficos, cloud computing
Abstract: In the field of cloud computing, lack of same ca pacity for devices as well as the lack of clear
knowledge on the number of used devices lead to not use of practical algorithms easily, so using optimal
algorithm is proper when any device fails to go beyond its maximum ca pacity. Also, in this paper, the
stability of partitioning has been considered. In fact, a subgraph must be selected in ter ms of connectivity
so that if a number of links are disconnected, its stability will not be missed and the program will run
correctly. In this paper, a method was presented for partitioning the task graph of an applic ation. Since the
problem of large -scale graph p artitioning is of the NP-hardnes s, the genetic algorithm was proposed as a
selective structure in the proposed method. In the genetic algorith m, three criteria including cost, response
time, and energy were used as a combined target. Use of linear programming e ffectively influences the
optimal performance of the genetic algorithm. The results of the p roposed method show accurate
reduction in energy consumption and the response time by 0.5% and 3%, respectively.
KEYWORDS: Genetic algorithm, linear programming, graph partitioning, cloud computing
1. INTRODUCTION
Today, the extension of knowledge
boundaries is dependent on the development of
computational technologies. As a starting point
for these technologies, we can refer to computer
network emergence in which only a few
computers were co nnected. Subsequently, these
small networks were connected to each other and
created the Internet that the networks were
shared on the Internet by which the concept of a
global spreadsheet was created through which
information was shared a mong users. Cloud
computing is a new way of providing computing
resources and a model for providing service over
the Internet. In fact, cloud computing provides
the ability to save on IT resources and boost
computing power, so that processing power
becomes an ever-reaching tool.
Mobile cloud computing is a combination
of cloud computing, mobile computing and
wireless networks that provide rich computing
resources for mobile users, network operato rs,
and cloud computing providers. The main
purpose of mobile cloud computing is to run
powerful mobile applications on a large number
of mobile devices. . In the mobile cloud
computing, processing is done in the cloud, data
is also stored in the cloud, and mobile serves as a
medium for displaying the information. With
cloud computing maturing, mobile cloud loading
has been become a hop eful way to reduce t he
runtime and battery life of mobile devices. Its
main idea is to boost the run by transferring
heavy computing from mobile devices to cloud
servers, and then getting results from them
through wireless networks. Loading is an
effective way to overcome the limited resources
and features of mobile phones since it can rid
them of massive co mputing and increase the
performance of mobile applications. Loading all
of the computing components of an application
remotely is not always necessary o r effective. In
mobile cloud computin g, the cell phone should
intelligently determine which part of the software
should be computed on the mobile or loaded into
the cloud. When mobile computing is
increasingly interacting with t he cloud, a number
of methods, for example, Maui [1] and Clone
Cloud [2], are provided with the aim of loading
some parts of the mobile software in the cloud.
In order to achieve a good performance, it must
be decided that which parts of the so ftware
should be loaded into a remote cloud, and which
parts of that should be run locally on mobile
devices, so that the total execution cost have
been be minimized. The main costs for mobile
loading s ystems are the cost of local computing
and re mote execution, and communicatio n costs
mean additional communication between the
mobile device and the cloud r emotely. The
computing can naturally be described as a graph
whose crest represents computational costs, and
its edge reflects communication costs [3]. By
partitioning the vertices of a graph, we can
divide computing among the local mobile
processors and remote cloud servers. The
traditional algorithm for gra ph partitioning (for
example, [4], [5], [6], and [7]) cannot be used
directly in mobile loading systems, because they
ignore the weight of each node and only consider
the margin weight of the graphs.
Cloud computing is the deepest distributed
computing technology that is rapidly penetrating
in various aspects of hum an life, aimed at
providing easy access to integrated clusters and a
unit of computing resources (stor age, processing
and storage) on demand, which can be dispersed
as conditional elastic with minimal investment,
complexity and interaction. [8] . Users use the
computing and resources sto red in the cloud, and
It is possible to access them at any time and
place by any device connected to the Internet
without the need for additional costs or condition
and, for example, the photos, songs and personal
documents in the cloud is available without any
time or place restrictions [9].
Mobile cloud computing is a technology
that directs diverse clouds and network resources
toward unlimited capability, storage and access
without considering the time and space for many
mobile devices by the Internet [ 10]. Cloud
computing brings the user's attention to integral
computing by reducing or eliminating confusion,
unusual actions, inaccurate and out-of-frame
results [11] , which provide s the users of mobile
cloud co mputing with greater benefits over
mobile computing and cloud computing. [12]
Researchers use diverse design patterns in
mobile c loud computing to design and optimize
cloud resources o n smartphones as a suggested
framework for processing distributed
applications on smartphone [13 ] for example,
clone cloud [14].
2. MATERIALS AND METHODS
Para continuar leyendo
Solicita tu prueba