International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 1765

ISSN 2229-5518

Issues related to Resource Management and Scheduling in Grid Computing and a Review of some Resource Management Models and

Harwinder Kaur, Student, Department of CSE Lovely Professional University

Abstract— continuously needed updating in better and challenging application programs via computational power have given birth to another need of connecting sky-high performance computational channels sources spreader along many organizations. So that’s why the core of grid computing is known as resources management. Many different work and projects have been centralizing on the praiseworthy implementation of resources management system with some collective choice of architecture and taxonomy. Nevertheless, the management of resources and scheduling computations in the grid environment is a multi-mix constructed project as they are equally divided, miscellaneous in mature, in hand of different single or multi organizations while having their different company policies, rules and regulations. Different access and cost models, and have dynamically varying loads and availability needs a powerful resource management system. This delivers quite count of challenging matter like site autonomy, heterogeneous substrate, policy extensibility, resource allocation or co-allocation, online control, scalability, transparency, and “economy of computations”. This paper will locate and talk about issues in resources management and scheduling in the emerging grid computing context and review of some resource management systems and scheduling techniques.

Index Terms— Grid Computing, Computational Grid, Resource Management, Nimrod/G, CONDOR G, Resource Broker, Scheduling, Quality of ser- vice (QOS)

.

1 INTRODUCTION

—————————— ——————————
he moral of grid computing is to earn favoured fame with the growth of internet as a planetary media and magnify- ing computational power needs of grand demanding ap-
plications. A computational grid [1] which is interest in cou- pling geographically distributed resources is also expanding in correcting wide-scale problems.
Computational grids are expected to offer dependable, con- sistent, pervasive, and inexpensive access to high-end re- sources irrespective of their physical location and the position of access points [2]. Fault tolerance, stability, adaptability, ex- tensibility, scalability and some others [3] should be brought forward by resource management system and scheduling techniques, to make grid to backing up a variety of applica- tions. The system of resources that are accessible to the grid are administrated by RMS. RMS is helping by keeping the trust of all resources providers, because it is require to do so as pool in a grid can have resources for few different providers. Application may request resources from the grid either pri- marily or secondary way. These can be assigned as jobs from the grid. For this it declares that RMS which is requiring per- forming resources management decision during it maximizes the QOS (quality of service) passed to the separate clients. The main goal of this paper is to highlight various issues related to resource management system and scheduling system for the success of grid computing and it also represent review of some resource management model and scheduling. Schedul- ing is appears simple, but complexity arises when users place QOS constraints like execution time and computation cost lim- itations. Such a guarantee of service is hard to provide in a grid environment as its resources are shared, heterogeneous,
distributed in nature, and owned by different organisations having their own policies and charging mechanisms [10]. In addition, scheduling algorithms need to adapt to the changing load and resource availability conditions in the grid in order to achieve performance and at the same time meet cost con- straints. 2 Procedure for Paper Submission

2.1 Review Stage

The term Grid was coined in the late 90s [1] in order to de- scribe a set of geographically distributed resources. The Grid is analogues to electrical power grid: access to computation and data should be as easy, pervasive, inexpensive and stand- ard as plugging in an appliance into an outlet [4]. There are number of resource management architectures have been pro- posed at the Grid Forum (GF) [5] Scheduling Working Group that are implemented, and deployed mainly based on three architectural models [6]: hierarchical, abstract owner, and market. Some of the Grid Resource Management systems are: The application level scheduling (AppLeS) [7] project, which has the main focus is to develop scheduling agents for indi- vidual applications on production computational Grids. Due to the focus of AppLeS is on scheduling, it follows the re- source management model supported by the underlying Grid middleware systems. AppLeS scheduler does not offer QOS support.
The Condor [6] environment follows a layered architecture and the both sequential and parallel applications are support- ed by it. The Condor system allocates the resources in the condor pool according to the usage conditions defined by re- source owners. There are multiple Condor pools that follow the flat RMS organizations. While the main focal point of

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 1766

ISSN 2229-5518

Condor software tools is harnessing the power of opportunis- tic and dedicated resources, Condor-G [6] is a derivative soft- ware system, which weights the software from Condor and Globus with main focus of job management services for grid applications. This isbasically an integration of interdomain resource management protocols of Globus (GRAM, Index Ser- vices) and the intradomain resource management methods of Condor. The DataGrid project is a multi-tier hierarchical RMS organization. It has a protractible schema based resource model with a hierarchical namespace organization. It has no QOS support.
With the participation of 11 partners and 6 European Union countries The European Commission has started project EU- ROGRID [8] which is a shared-cost Research and Technology Development project, for the formation of an international network of high performance computing centres.
The Netsolve system [6] that is a computational Grid based on hierarchical machine organization. Basically, Netsolve agent’s reatin the log about assorted resources available in the net- work. And it is the accountability of the Netsolve servers Grid for making their existence aware to Netsolve Agent and make use of push protocol for resource dissemination. These agents are also archieve for the task of resource discovery and sched- uling.
A global resource management and scheduling system for computational grid named, Nimrod/G resource broker, built using Globus services [2]. The main supremacy of Nimrod/G is that it supports deadline and cost-based scheduling mecha- nism, but the costing mechanism is in static mode. The Globus metacomputing toolkit does not provide any type of services for dynamically transaction of resources. This constraint of Globus is beaten by middleware infrastructure called GRid Architecture for Computational Economy (GRACE) that coex- ist with Globus, and Nimrod/G can use for dynamically transaction of resources. Because of the geographically alloca- tion of multi organizational resources, The management for resources and scheduling system in the grid computing envi- ronment needs address various issues such as: availability, scalability etc, are needed to resolve. These issues are high- lighted in table 1.
In 2000, Buyya and Giddy proposed market/economy model based architecture for Grid resource management [10]. The main key components of economy -driven resource manage- ment system proposed by Buyya include: User Applications (sequential, parametric, parallel, or collaborative apps),DDThe Grid Resource Broker (a.k.a., Super/Global/Meta Sched- uler),DDGrid Middleware, The Domain Resource Manager (Local Scheduler or Queuing system) [10].
In 2002, junwei cao, Stephen proposed an agent based re- source management system for grid (ARMS) [9] that address two key issues of resource management. ARMS [9], is basically put into practice for grid computing. The main modus op- erandi utilises by this model is the performance prediction techniques of the PACE [9] toolkit in order to provide quanti- tative data concerning the performance of composite applica- tions running on a local grid resource.A hierarchy of homoge- neous agents are used to provide a scalable and adaptable ab- straction (which are the two main key isssues covered by this
model) at the meta level of the system architecture.Each and every agent is able to assist with other for providing service advertisement and discovery of resources for the scheduling of applications that need to exploit grid resources.
The University of California at San Diego (UCSD), developed a toolkit named, SIMGRID [11], which is C based toolkit for the imitation of application scheduling. It supports modelling of time-shared resources and the constant load can be injected or from real traces. This system allows the formation of requi- sites in terms of their execution time and resources with re- spect to a standard machine capability which makes it a dom- inant system.
In 2003, R.Buyya implemented GridSim [12] in Java by lever- aging SimJava’s [12] basic discrete event simulation infrastruc- ture. The feature of Gridsim implemented in java is likely to request to educators and students since Java is one of the pop- ular programming language now’s day for network compu- ting. Salient features of the GridSim [12] toolkit include - mod- elling of heterogeneous types of resources, modelled the re- sources under space- or time -shared mode, and potentially defined the resources.

TABLE 1

ISSUES RELATED TO RESOURCE MANAGEMENT SYSTEM AND

SCHEDULING

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 1767

ISSN 2229-5518

there are many typical matters also related to these two as well. It can be corrected in time and budgeted money by indi- vidual resource or organisation rather than a large count of applications which need more computing power. By having all this, it helped in finding of logically coupling geographical- ly distributed high-end computational resources and applying them to solve using them for solving good number of issues.

REFERENCES

[1] Foster, I., & Kesselman, C. (Eds.). (2003). The grid 2: Blueprint for a new computing infrastructure. Mor- gan Kaufmann.

[2] Buyya, R., Abramson, D., & Giddy, J. (2000, June). An economy driven resource management architecture for global computational power grids. In Proceedings of the 2000 International Conference on Parallel and Distributed Processing Techniques and Applications (PDPTA 2000) (pp. 26-29).

[3] Krauter, Klaus, Rajkumar Buyya, and Muthucumaru Maheswaran. "A taxonomy and survey of grid re- source management systems for distributed compu- ting." Software: Practice and Experience 32.2 (2002):

135-164.

[4] I. Foster. The Grid: A New Infrastructure for 21st Cen- tury Science. Physics Today, 55(2):42, February 2002.

[5] Grid Forum - http://www.gridforum.org/

[6] Buyya R., Chapin S., DiNucci D., Architectural Mod- els for Resource Management in Global Computa- tional Grids, http://www.buyya.com/ecogrid/

[7] Berman, F., & Wolski, R. (1997, May). The AppLeS project: A status report. In Proceedings of the 8th NEC Research Symposium (Vol. 16). Berlin, Germany, May.

[8] Joseph, J., & Fellenstein, C. (2004). Grid computing.

IBM press.

[9] Cao, J., Spooner, D., Turner, J. D., Jarvis, S., Kerbyson, D. J., Saini, S., & Nudd, G. (2002, May). Agent-based resource management for grid computing. InCluster Computing and the Grid, 2002. 2nd IEEE/ACM Interna- tional Symposium on (pp. 350-350). IEEE.

3 CONCLUSION

As many hand work, project approaches and architectures available for resource management and scheduling system,

[10] Buyya, R., Giddy, J., & Abramson, D. (2000). An eval- uation of economy-based resource trading and scheduling on computational power grids for param-

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 1768

ISSN 2229-5518

eter sweep applications. Active Middleware Services,

221-230.

[11] Casanova, H. (2001). Simgrid: A toolkit for the simula- tion of application scheduling. In Cluster Computing and the Grid, 2001. Proceedings. First IEEE/ACM Inter- national Symposium on (pp. 430-437). IEEE.

[12] Buyya, R., & Murshed, M. (2003). Gridsim: A toolkit for the modeling and simulation of distributed re- source management and scheduling for grid compu- ting. Concurrency and Computation: Practice and Experi-

ence, 14(13‐15), 1175-1220.

[13] Kertész, A., & Kacsuk, P. (2009). Grid Interoperability Solutions in Grid Resource Management. Systems Journal, IEEE, 3(1), 131-141.

IJSER © 2013 http://www.ijser.org