ENHANCED GRID RESOURCE SELECTION MECHANISM

Adil Yousif, Abdul Hanan Abdullah

Faculty of Computer Science and Information System, Universiti Teknologi Malaysia, Johor, Malaysia

Muhammad Shaﬁe Abd Latiff, Mohammed Bakri Bashir

Faculty of Computer Science and Information System, Universiti Teknologi Malaysia, Johor, Malaysia

Keywords:

Grid computing, Resource selection, Resource management and bidding.

Abstract:

Resources in grid systems are heterogeneous, geographically distributed, belong to different administrative

domains and apply different management policies. To manage resources in grid computing, efﬁcient resource

selection mechanism is required. In traditional non-reserved resource selection mechanisms, there is no guar-

antee that the job completion time will be as expected. On the other hand, in the reserved bidding process,

conventional resource selection mechanisms waste providers resources and hence affect the overall grid sys-

tem performance. We propose a new grid resource selection mechanism to achieve performance enhancement

in the resource allocation process.

1 INTRODUCTION

Grid computing emerged in the middle of 1990s as a

wide-scale distributed system to offer dynamic coor-

dinated resources sharing and high performance com-

puting (Foster and Kesselman, 2004). Grid technolo-

gies have evolved over the past two centuries from

primary metacomputing into Open Grid Service Ar-

chitecture(OGSA) using the service oriented architec-

ture (SOA) Concepts.

Grid computing as deﬁned previously differs with

current internet technologies, as the current inter-

net technologies handle the process of exchanging

data between internet users, but they do not sup-

port methods for coordinated employment of several

resources at different internet locations for compu-

tation (Iamnitchi and Foster, 2001). CORBA and

Enterprise Java as distributed computing technolo-

gies differ with grid computing as they allow shar-

ing of resources inside one institute or company but

not with other outer institutes or companies (Harold

and Loukides, 2000; Iamnitchi and Foster, 2001).

In Storage Service Providers(SSP) and Application

service Providers(ASP), Institutes are permitted to

serve outside clients but this occurs in restricted man-

ners, such as using special types of network connec-

tions (Iamnitchi and Foster, 2001). According to the

previous review current technologies do not handle

all resource categories or do not support the manage-

ment and ﬂexibility needed to build the coordinated

resource sharing among several institutes and organi-

zations.

This paper is organized as follows: Section 2 de-

scribes the resource management in grid computing;

this is followed, in Section 3, by a discussion of the

related works in the area. Details of the proposed

mechanism are presented in Section 4; followed by

a conclusion and discussion.

2 RESOURCE MANAGEMENT IN

GRID COMPUTING

2.1 Overview

Resources in grid systems are heterogeneous, geo-

graphically distributed, belong to different adminis-

trative domains and apply different management poli-

cies (Ahmed et al., 2007; Chang et al., 2011). Further-

more, grid management systems do not have full con-

trol over the resources that belong to the grid (Schni-

zler, 2007). Czajkowski et al. (2002) deﬁned the term

resource to denote any capability that may be shared

and exploited in a networked environment. The re-

sources and services may differ in the form of func-

tionality that they offer to the user, but both are sim-

ilar in the way that they provide the functionalities

226

Yousif A., Hanan Abdullah A., Shaﬁe Abd Latiff M. and Bakri Bashir M..

ENHANCED GRID RESOURCE SELECTION MECHANISM.

DOI: 10.5220/0003606202260229

In Proceedings of the 6th International Conference on Software and Database Technologies (ICSOFT-2011), pages 226-229

ISBN: 978-989-8425-76-8

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: Grid Resource Broker.

to the users (Foster and Kesselman, 2004; Andrieux

et al., 2004); hence we can use the term resource more

generally to indicate all types of ancient resources, as

well as services.

In conventional computing systems, such as PCs,

cluster systems and other high performance comput-

ing, resource management problem has been stud-

ied intensively and a number of different types of

management systems are implemented such as batch

scheduler, workﬂow engine and local operating sys-

tems. These systems have full control over resources

that follow them and also have a full understanding of

all information on those resources (Schnizler, 2007).

The local recourse management systems are not capa-

ble of managing grid computing resources since the

heterogeneity of resources and environments in grid

networks. Moreover, grid management systems do

not have the full control over the resources that they

manage (Schnizler, 2007). Resource management in

grid environments is a great challenge; this is due to

the heterogeneity of resources in grid environments

and besides that, the resources belong to diverse ad-

ministrativedomains and apply different management

policies (Wang et al., 2010).

2.2 Resource Brokers

In the decentralized resource broker systems, each

grid user has its own broker to access the avail-

able grid resources. All resources that participate in

the grid system must register in one or more index

servers. However, these index servers have the abil-

ity to register in other index servers. The discovery of

available grid resources is made by the user’s resource

brokers through communicating between client bro-

ker with one or multiple index servers and the result

of this process is a list of all available resources that

satisﬁed the user needs. If the resource broker needs

more information, it can contact the resource provider

directly. Before the submission of the user jobs, the

user broker select suitable resources from the avail-

able resources depending on some selection criteria

such as the Quality of service (QoS) requested by

the user, and locations and policies of the resource

providers (Krauter et al., 2002; Elmroth and Tords-

son, 2008). Figure 1 describes decentralized resource

broker process.

2.3 Co-allocation and Advance

Reservation

In some scenarios grid clients request more than one

resource at the same time to perform a speciﬁc task,

part of those resources is meaningless to the client.

Accordingly, the resource allocation system should

ﬁnd a way to provide those resources all together

as a bundle. The process for requesting a number

of homogeneous resources concurrently by the grid

clients is called Co-Allocation; therefore, the alloca-

tion manager should support a co-allocation mecha-

nism (Schnizler, 2007; Elmroth and Tordsson, 2009).

Advance reservation is an important issue arise

when grid clients request grid resources at a certain

time in future; for instance, a grid client demands pro-

cessor after two hours, and in some cases the grid

client needs resources for a certain period of time

ENHANCED GRID RESOURCE SELECTION MECHANISM

227

Figure 2: Matchmaking Approach.

in the future, such as a client requested memory af-

ter two hours and for one hour. Some grid clients

need resources in a time in the future, but the time

is not well-deﬁned; for instance, a grid client needs

resources after two or three hours. Therefore, the al-

location management system must comprise a way for

the grid clients to describe the time in which they need

the resources precisely; and this may also help the

resource providers to clarify at which times their re-

sources will be idle in the future (Elmroth and Tords-

son, 2008, 2009) .

3 RELATED WORK

Due to the heterogeneity of grid resources, there is

a need to describe those resources in a standard way

so that providers and grid clients can allocate them

in consistent way (Park and Kang, 2007; Yuesheng

et al., 2009). There are two types of resource se-

lection approaches for grid environments, matchmak-

ing approach and bidding based approach. In match-

making approach (Figure 2) resource providers de-

scribe and register their resources in the matchmaker

database. When a grid client requests resources to

perform tasks, it determines the properties of those

resources to the matchmaker (Yuesheng et al., 2009);

and the matchmaker will perform a matching algo-

rithm to match the resources to tasks and sends noti-

ﬁcation to the grid client and to the selected resource

provider (Wang et al., 2010). The resource providers

update their resources states to the matchmaker peri-

odically and any changes occur between the updating

processes is not reﬂected in the matchmaker. More-

over, the matchmaker like other centralized systems

represents a performance bottleneck and single point

of failure (Wang et al., 2010).

The idea behind bidding based approach arose to

avoid the matchmaking drawbacks. In the bidding

process when a grid client needs resources to execute

certain jobs, it forwards a call for proposal CFP mes-

sage to grid providers (Hongbo et al., 2005; Wang

et al., 2010). Upon receiving the CFP messages, re-

source providers decide whether or not to join the

bidding process according to their facilities and re-

sources characteristics. If a grid provider partici-

pated in a bidding process it dispatches a bid to the

client. The client assesses all the received bids and

chooses the best bid according to a speciﬁc bidding

algorithm (Habibi and Jafar, 2009; Vilajosana, 2008;

Wang et al., 2010).

Resource securing in grid resource allocation pro-

cess can be achieved in two different ways; In re-

served securing and non-reserved resource securing

(Wang et al., 2010). In reserved method the resource

providers keep the resources for the grid client’s bids;

so it can guarantee the future status of the resources

(Haji et al., 2006). However, this method wastes re-

sources because more than one resource provider may

keep their resources for a single bid and at the end the

grid client will select only one resource and reject the

others; and so those rejected resources are wasted at

the time that they have been kept for that client’s bid

(Wang et al., 2010).

On the other hand, in non-reserved resource secur-

ing method the resource providers do not keep the re-

sources for any bid; and hence the resource providers

can utilize their resources effectively; nevertheless,

this method does not promise the future status for the

resources; and hence there is no guarantee that the job

completion time will be as expected (Haji et al., 2005;

Yousif et al., 2011).

4 THE PROPOSED MECHANISM

We propose a grid resource selection technique to

tackle the problems of conventional reserved and non

reserved bidding approaches; following is a brief de-

scription of our mechanism. In our mechanism when

a grid client requests resources, it create a call-for-

proposal CFP messages and forwards it to all avail-

able resource providers in the grid system and wait

for a certain time interval for bids. Upon receiving

the CFP messages, resource providers decide whether

or not to join the bidding process according to their

facilities and the resource’s characteristics. If a re-

source provider participated in a bidding process it

dispatches a bid to the client and reserves the corre-

sponding resources for that bid. The bids arrive at the

grid client one by one; and the clients have to assess

the received bids and make judgments immediately.

At the beginning the ﬁrst bid arrives at the grid client

is considered as the best bid; and when the next bid

arrives at the client, the client compares the current

received bid with the best bid, if the current received

bid is better than the best bid the client will consider

ICSOFT 2011 - 6th International Conference on Software and Data Technologies

228

the current received bid as the best bid and notiﬁes the

ﬁrst resource provider that its bid is rejected and then

the provider will free the resources. However, if the

old best bid is better than the current received bid the

grid client simply sends reject notiﬁcation message to

the new provider to free its resources.

5 CONCLUSIONS

This paper presents a new resource selection mecha-

nism in a bidding based grid system to enhance the

process of resource selection in an effective and efﬁ-

cient way. The proposed mechanism tackles the prob-

lems arise when using traditional reserved and non-

reserved resource selection mechanisms. Our algo-

rithm keeps the most suitable resource for executing

the job as a promise, and therefore, the job comple-

tion time will be as estimated. Furthermore, since

the proposed mechanism reserves only one resource,

resource providers can utilize their resources effec-

tively. Now we are in the implementation phase, and

we plan to simulate the mechanism in a grid simulated

environmentusing GridSim as a simulator, to evaluate

the performance of the proposed mechanism.

REFERENCES

Ahmed, A., Latiff, M., Bakar, K., and Rajion, Z. (2007).

Adaptive resources selection framework for grid en-

abled visualization pipeline. IJCSNS, 7(12):114.

Andrieux, A., Czajkowski, K., Dan, A., Keahey, K., Lud-

wig, H., Pruyne, J., Rofrano, J., Tuecke, S., and Xu,

M. (2004). Web services agreement speciﬁcation (ws-

agreement).

Chang, R. S., Lin, C. F., and Chen, J. J. (2011). Selecting

the most ﬁtting resource for task execution. Future

Generation Computer Systems, 27(2):227–231.

Czajkowski, K., Foster, I., Kesselman, C., Sander, V., and

Tuecke, S. (2002). Snap: A protocol for negotiating

service level agreements and coordinating resource

management in distributed systems. pages 153–183.

Springer.

Elmroth, E. and Tordsson, J. (2008). Grid resource bro-

kering algorithms enabling advance reservations and

resource selection based on performance predictions.

Future Generation Computer Systems, 24(6):585–

593.

Elmroth, E. and Tordsson, J. (2009). A standards-based grid

resource brokering service supporting advance reser-

vations, coallocation, and cross-grid interoperability.

Concurrency Computation Practice and Experience,

21(18):2298–2335.

Foster, I. and Kesselman, C. (2004). The grid: blueprint for

a new computing infrastructure. Morgan Kaufmann.

Habibi, B. N. and Jafar (2009). Bidding strategically for

scheduling in grid systems. Journal of Information

Processing Systems, 5(2):87–96.

Haji, M., Djemame, K., and Dew, P. (2006). Deployment

and performance evaluation of a snap-based resource

broker on the white rose grid. volume 2, pages 3365–

3370. IEEE.

Haji, M. H., Gourlay, I., Djemame, K., and Dew, P. M.

(2005). A snap-based community resource broker us-

ing a three-phase commit protocol: A performance

study. Computer Journal, 48(3):333–346.

Harold, R. and Loukides, M. (2000). Java network pro-

gramming. O’Reilly & Associates, Inc. Sebastopol,

CA, USA.

Hongbo, Z., Hai, J., Zongfen, H., Jing, T., and Xuanhua,

S. (2005). A virtual-service-domain based bidding al-

gorithm for resource discovery in computational grid.

In Web Intelligence, 2005. Proceedings. The 2005

IEEE/WIC/ACM International Conference on, pages

50–53.

Iamnitchi, A. and Foster, I. (2001). On fully decentralized

resource discovery in grid environments. Grid Com-

putingGRID 2001, pages 51–62.

Krauter, K., Buyya, R., and Maheswaran, M. (2002). A tax-

onomy and survey of grid resource management sys-

tems for distributed computing. Software: Practice

and Experience, 32(2):135–164.

Park, J. and Kang, J. (2007). Optimization of xquery queries

including for clauses. page 37. IEEE.

Schnizler, B. (2007). Resource Allocation in the Grid: A

Market Engineering Approach. Univ.-Verl. Karlsruhe.

Vilajosana, X. (2008). Bidding support for computational

resources. pages 309–314. IEEE.

Wang, C.-M., Chen, H.-M., Hsu, C.-C., and Lee, J.

(2010). Dynamic resource selection heuristics for a

non-reserved bidding-based grid environment. Future

Generation Computer Systems, 26(2):183–197.

Yousif, A., Abdullah, A. H., and Ahmed, A. A. (2011). Ar-

ticle: A bidding-based grid resource selection algo-

rithm using single reservation mechanism. Interna-

tional Journal of Computer Applications, 16(4):39–

43. Published by Foundation of Computer Science.

Yuesheng, T., Lu, H., and Jingyu, W. (2009). Research

of the xml-based personal resource description in grid

computing. volume 2, pages 639–641. IEEE.

ENHANCED GRID RESOURCE SELECTION MECHANISM

229