Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more ➡
Download
Standard view
Full view
of .
Add note
Save to My Library
Sync to mobile
Look up keyword
Like this
3Activity
×
0 of .
Results for:
No results containing your search query
P. 1
A Hierarchical Resource Allocation Architecture for Mobile Grid Environments

A Hierarchical Resource Allocation Architecture for Mobile Grid Environments

Ratings: (0)|Views: 1,232|Likes:
Published by ijcsis
The mobility issue in grid environments has established new challenges to the research communities particularly in the areas of scheduling, adaptation, security and mobility. Especially, the resource allocation becomes more challenging when mobility is considered in grid environment. Hence it is necessary to consider the mobility of users along with the resource availability while scheduling the resources for the execution of jobs. In this paper, we propose to design a Hierarchical Resource Allocation Architecture (HRAA) which includes resource monitoring and scheduling operations for mobile grid. In this architecture, the Mobile Grid is divided into clusters. Each cluster has one cluster head (CH). A master server (MS) controls each local clusters and has frequent updates of all the CH information. Each CH has a monitoring agent (MA) which will periodically predict the mobility of the cluster nodes and monitor the resource availability and update their values. When the MS forwards the job request of a user to the ideal CH, the CH schedules the jobs based on the predicted time for resource availability and sufficiency of the resources. By simulation results, we show that our proposed architecture achieves good success ratio and throughput with reduced delay and energy consumption.
The mobility issue in grid environments has established new challenges to the research communities particularly in the areas of scheduling, adaptation, security and mobility. Especially, the resource allocation becomes more challenging when mobility is considered in grid environment. Hence it is necessary to consider the mobility of users along with the resource availability while scheduling the resources for the execution of jobs. In this paper, we propose to design a Hierarchical Resource Allocation Architecture (HRAA) which includes resource monitoring and scheduling operations for mobile grid. In this architecture, the Mobile Grid is divided into clusters. Each cluster has one cluster head (CH). A master server (MS) controls each local clusters and has frequent updates of all the CH information. Each CH has a monitoring agent (MA) which will periodically predict the mobility of the cluster nodes and monitor the resource availability and update their values. When the MS forwards the job request of a user to the ideal CH, the CH schedules the jobs based on the predicted time for resource availability and sufficiency of the resources. By simulation results, we show that our proposed architecture achieves good success ratio and throughput with reduced delay and energy consumption.

More info:

Published by: ijcsis on Sep 05, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See More
See less

02/28/2011

pdf

text

original

 
A Hierarchical Resource Allocation Architecture forMobile Grid Environments
S. Thenmozhi
Assistant Professor, Department of Computer ApplicationsChettinad College of Engineering and TechnologyTamilnadu, India.sthenmozhiphd@gmail.com
A.Tamilarasi
Professor, Department of MCAKongu Engineering College, ErodeTamilnadu, India
 Abstract—
The mobility issue in grid environments hasestablished new challenges to the research communitiesparticularly in the areas of scheduling, adaptation, security andmobility. Especially, the resource allocation becomes morechallenging when mobility is considered in grid environment.Hence it is necessary to consider the mobility of users along withthe resource availability while scheduling the resources for theexecution of jobs. In this paper, we propose to design aHierarchical Resource Allocation Architecture (HRAA) whichincludes resource monitoring and scheduling operations formobile grid. In this architecture, the Mobile Grid is divided intoclusters. Each cluster has one cluster head (CH). A master server(MS) controls each local clusters and has frequent updates of allthe CH information. Each CH has a monitoring agent (MA)which will periodically predict the mobility of the cluster nodesand monitor the resource availability and update their values.When the MS forwards the job request of a user to the ideal CH,the CH schedules the jobs based on the predicted time forresource availability and sufficiency of the resources. Bysimulation results, we show that our proposed architectureachieves good success ratio and throughput with reduced delayand energy consumption.
 Keywords-Resource Allocation, monitoring agent (MA), Hierarchical Resource Allocation Architecture (HRAA), cluster head, master server (MS)
I.
 
I
NTRODUCTION
 
 A. Mobile Grid 
The Grid is a distributed, high performing computing anddata handling infrastructure. It provides common interfaces forall the resources by using standard, open, general-purposeprotocols and interfaces by incorporating the geographicallyand organizationally dispersed, heterogeneous resources. Butit is the basis and the enabling technology for the persistentand utility computing [1]. Multiple administrative domains,autonomy, heterogeneity, scalability anddynamicity/adaptability are the important features of the Grid.The mobility issues are handled by enabling both fixed andmobile users, in the mobile grid environment. By using theunderlying technologies transparently and efficiently, theaccess for both fixed and mobile grid resources are provided.Mobile Grid is derived from Grid with the additional supportof mobile users and resources in a seamless, transparent,secure and efficient way. Moreover, it has the capability toorganize the underlying ad hoc networks. It forms arbitraryand unpredictable topologies by providing a self-configuringgrid system of mobile resources which are connected by thewireless links [1].The mobile grid uses the advanced capabilities of wirelessnetworks and lightweight thin devices. Though grid computingintegrates geographically dispersed resources and users tocreate a dynamic virtual organization, most of the resourcesare static in nature. The user and the resource participating inproblem solution are the two basic units of the processingenvironment. [2].For many large scale applications which are dynamic innature and require transparency for users, Mobile Grid isconsidered as the best solution. Grid will increase the jobthroughput and performance of the corresponding applicationsby applying efficient mechanisms for resource management.Moreover, it will enable the advanced forms of cooperativework by allowing the seamless integration of resources, data,services and ontologies [1]. Some of the applications of themobile grids are scientific, public services and commercialbusinesses. Mobile grids integrate the mobile devices likelaptops, PDAs (Personal Digital Assistants) [2].
 B. Resource Allocation (or) Management in Mobile Grid 
Resource allocation is a basic issue to achieve highperformance on a grid workflow [3]. Resource allocation canbe classified into resource selection (discovery) and resourcebinding (acquiring). The resource selection is separated fromresource binding based on the common architecture of conventional resource brokering systems. These systemsmainly concentrate on the resource selection for providingcomplex resource specification languages and resourceselection algorithms. On providing a resource specification byusing available resource information, a resource selectionalgorithm first discovers a matching set of resources andnegotiates with an individual local resource manager. Then theapplication attempts to acquire the resources [4].
C. Resource Allocation challenges in Mobile Grid 
When mobility is considered, the resource selectionbecomes more challenging. Therefore, it is necessary toconsider the mobility of users with the resources in resource
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 20106http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
selection. The mobility issue in grid environments hasestablished new challenges to the research communitiesparticularly in the areas of scheduling, adaptation, security andmobility. Mostly, the behavior of the user and/or mobiledevice is highly unpredictable which produces disconnectionproblemsIn grid environments, new challenges are introduced to theresearch communities especially in the fields of mobility,scheduling, adaptation and security due to issues like mobility,power consumption and size of devices. The secondaryproblems are the small screen size and difficult inputmechanisms. Peer-to-peer computing provides many usefultechnologies and ideas, for developing scalable and reliablemobile grids. In mobile environment, the most challengingproblem is the disconnection problem [2].The problems of resource allocation are:1)
 
Identification of an appropriate service and resources.2)
 
Based on certain criteria, allocating the resources,such as pricing or priority.3)
 
Dynamically allocating and updating the state of theresources [5].A centralized allocation manager is not possible becausethe portions of the Grid may apply different allocationstrategies due to decentralized Grid policies. The lack of accurate resource status information at the global scale is theadditional challenge to the Grid resource allocation. At theirremoval, the knowledge of real time environment has beenlimited by the allocation strategies which are utilized by theusers and brokers. Therefore the possible allocationmechanisms should not depend on the availability of currentglobal knowledge [6].In this paper, we develop a resource monitoring andscheduling scheme for mobile grid. In this scheme, the usersubmits the job request for a job to be executed to a server.The job request contains job description, number of resourcesrequired, expected job completion time and the quoted priceallotted for it.II.
 
R
ELATED
W
ORK
 Xiaozhi Wang et al [7], have proposed a layered structureof Grid QoS. Based on the analysis of the content of gridresource allocation management (GRAM) based on QoS, theirwork puts forward the architecture of GRAM based on QoS.Through mapping, converting and negotiating the QoSparameters, it can implant the user's requirement about QoS inthe process of resource allocation management, and connectGrid QoS with GRAM very well. All these provide areasonable consulting model for QoS and resource allocationmanagement in grid. Their work raised the performance of GRA from different aspects. In their process of searching,system may negotiate with the user, then get final result: notbeing able to supply, being able to supply or reducing QoSdemands to supply.Konstantinos Katsaros et al [8] have discussed a campus-wide hierarchical Mobile Grid system architecture in whichmobile nodes (MNs), willing to offer their computationalresources, move between WLANs. This willingness of themobile node is based on reciprocity criteria. They have alsoconsidered the divisible load applications (DLA) whichdivides the load of computation into parts and made to carryout independently. Moreover, they have described anarchitecture for the realization of a Mobile Grid andinvestigate key design decisions and optimizations.Lei Zhang et al [9] used the Particle Swarm Optimization(PSO), the latest evolutionary optimization technique to solvethe task scheduling problem in grid (computational grids)environment. Here each particle is represented as a possiblesolution, and the position vector is transformed from thecontinuous variable to the discrete variable. They have alsoaimed to generate an optimal schedule to get the minimumcompletion time while completing the tasks.Abdullah M. Elewi et al [10] have addressed the problemof energy efficient real time task scheduling where the tasksare dependent due to exclusive access shared resources.Moreover, they have proposed about the enhancements overthe existing dual speed switching algorithm (DSA) where theirproposed algorithm achieves more energy saving and has thecapability to function with both SRP and DPCP protocols.Homam Reda El-Taj et al [11] have given a survey aboutmobile computing, the mobile grid computing, mobile agentsand how to apply mobile agents on mobile grid computing andwhat has been done to solve the issues in these areas of study.Ming Wu et al [12] have proposed a prototype of GridHarvest Service (GHS) which provide dynamic and self-adaptive task scheduling. Their study is made upon task scheduling of a parallel or distributed application with adivisible workload in a heterogeneous environment. It alsoshows the possibility of integrating the three parts of task scheduling, that is the task allocator, scheduler and predictorinto existing toolkits for better service.Hesham A. Ali et al [13] have introduced a "self rankingalgorithm", which will be used to build a mobile computingscheduling mechanism to schedule the tasks on the mobiledevices, which will maximize the profit of the mobile deviceswhich are integrated within the grid using their computationalpower as an addition to the system overall power.Gurleen Kaur et al [14] have addressed the promising andbright side of the grid computing technology. They haveexplored the grid capabilities further by organizing the gridcomputing concept from two broad perspectives such asUser’s Perspective and Administrator’s Perspective.III.
 
H
IERARCHICAL
R
ESOURCE
A
LLOCATION
A
RCHITECTURE
 
 A. System Design
In our system design, a set of machines and a cluster head(CH) are included in each local cluster. A master server (MS)controls and groups many local clusters. The MS collectsinformation about the resources in its local clusters. Then itstores the information in its own database.The proposed scheme divides a given task that is submittedby a user into subtasks. Then it finds spare processors andother critical resources on the network, distributes the
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 20107http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
subtasks, monitors the progress of the subtasks, and restartsthe subtasks that fail.It consists of the following steps:1)
 
The MS divides a given task into a sequence of subtasks and allocates the subtasks to the local clusterheads (CH).2)
 
The CH finds the available processing power andaverage mobility on the local machines and thendistributes the subtasks over these machines.3)
 
The CHs gather the completed subtasks from the localmachines and then send the data back to the MS.4)
 
The MS aggregates the completed subtasks and thenstores the results in its own database, and5)
 
MS also sends back the results to the application of the respective users.
 B. Mobility Metric
User Range is a given fact about the initiator coverage, inwhich user can communicate with mobile devices. AverageMobility, a derived parameter, represents the average mobilityof a resource and/or user (based on user and resourcemobility). Average mobility is calculated based on two recentcommunications between user/initiator and resource withrespect to the user/initiator.Average Mobility is calculated as21
 f  f  Mobility  Average
=
(1)Where, 1
 f 
and 2
 f 
are the first history and second historyrespectively. The history is simply the distance between userand resource and it can be calculated by finding differencebetween the two recent interactions.
Ptime
shows the predicted time for resource availabilitywithin the user’s range and is calculated by the followingequation
Ptime
= (User Range Distance) / Average Mobility (2)The “Distance” is the net difference between the newlocations of user and resource.
C. Resource Availability Metric
The monitoring agents (MA) estimates the workload of itsgrid nodes ),,2,1,(
in
i
L
=
present in the cluster CL
1
usingthe following formula:
i j jii
Power  JobsizeCWLWL
    
+=
=
1
(3)Where
i
CWL
 
is the current workload of 
i
n
,
i
WL
is thework load of 
i
n
 
i
Power 
is the power of the node
i
n
and
 j
 Jobsize
is thesize of the
 j
 Job
.
 D. Scheduling Strategy
After estimating the
Ptime
and
WL
values, the MA sendsthese values to its cluster head
1
CH 
.
The
1
CH 
then schedulesthe jobs if their grid node satisfies the following condition
ThWLPtime If 
>
) / ( (4)where
Th
is a threshold value (which can be fixed basedon the job request).From 4, we can observe that if the
Ptime
is less and if thework load is more then the gird nodes are unable to executethe job request. Therefore, the jobs are executed by the nodesonly when
Ptime
is high and the work load is low. If the CHis unable to allocate the resources in its cluster, it resubmitsthe job request information to the MS.
 E. Functions of HRAA
Figure 1. Functions of HRAA
The sequence of operations in HRAA is shown in Figure. 1. Inthis figure, the arrows represent the communication messagesand the nodes represent the agents/servers. The sequence is asfollows.1)
 
The MAs of each node in the local cluster send theresource status information to the CHs.2)
 
The CHs send this information to the MS.3)
 
The MS then create a database which containsinformation about the status and the price of eachresource.4)
 
A user submits its job details and the resourcerequirements to the MS.5)
 
The MS sends the job request information to the localCH.6)
 
The local CH allocates the resources in their controldepending on the predicted time, the averageprocessing power and average load of its localmachines
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 20108http://sites.google.com/site/ijcsis/ISSN 1947-5500

Activity (3)

You've already reviewed this. Edit your review.
1 thousand reads
1 hundred reads
dhana_apk liked this

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->