Efficient Resource Management Mechanism with Fault Tolerant Model for Computational Grids

International Journal of Computer Applications Technology and Research
Volume 3– Issue 12, 774 - 777, 2014, ISSN:- 2319–8656
www.ijcat.com 774
Efficient Resource Management Mechanism
with Fault Tolerant Model for Computational
Grids
R. Kohila
Department of Computer Science and Engineering
V.S.B Engineering College
Tamilnadu, India.
Abstract- Grid computing provides a framework and deployment environment that enables resource
sharing, accessing, aggregation and management. It allows resource and coordinated use of various
resources in dynamic, distributed virtual organization. The grid scheduling is responsible for resource
discovery, resource selection and job assignment over a decentralized heterogeneous system. In the
existing system, primary-backup approach is used for fault tolerance in a single environment. In this
approach, each task has a primary copy and backup copy on two different processors. For dependent
tasks, precedence constraint among tasks must be considered when scheduling backup copies and
overloading backups. Then, two algorithms have been developed to schedule backups of dependent and
independent tasks. The proposed work is to manage the resource failures in grid job scheduling. In this
method, data source and resource are integrated from different geographical environment. Fault-
tolerant scheduling with primary backup approach is used to handle job failures in grid environment.
Impact of communication protocols is considered. Communication protocols such as Transmission
Control Protocol (TCP), User Datagram Protocol (UDP) which are used to distribute the message of
each task to grid resources.
Key Words: Grid Computing, Primary Backup, Communication Protocols, TCP-
Transmission Control Protocol, UDP- User Datagram Protocol.
1. INTRODUCTION
1.1 Grid Computing
Grid Computing is distributed; large
scale cluster grid computing has emerged as
the next-generation parallel and distributed
computing methodology, which aggregates
dispersed heterogeneous resources for solving
various kinds of large-scale parallel
applications in science, engineering and
commerce. It can integrate and utilize
heterogeneous computational resources from
different networks or regional areas into a high
performance computational platform and can
solve complex computing-intensive problems
efficiently. Grid service represents
convergence between high performance
computation and web service. Grid aims
ultimately to turn the global network of
computers into a vast computational resource.
1.2 Grid Computing Overview
A distributed heterogeneous
computing system consists of a distributed
suite of different high-performance machines,
interconnected by the high-speed networks, to
perform different computationally intensive
applications that have various computational
requirements. Heterogeneous computing
systems range from diverse elements or
paradigms within a single computer to a cluster
of different types of personal computers to
coordinate geographically distributed machines
with different architectures. Job scheduling is
one the major difficult tasks in a computational
grid.

www.ijcat.com 775
2. RELATED WORK
2.1 Scheduling
Lan Foster and Car Kesselman (2004)
[3] develop a fault tolerant job scheduling
strategy in order to tolerate faults gracefully in
an economy based grid environment. They
propose a novel adaptive task check pointing
based fault tolerant job scheduling strategy for
an economy based grid. They present a survey
with the grid community. The survey reveals
that, users have to be highly involved in
diagnosing failures, that most failures are due
to configuration problems and that solutions
for dealing with failures are mainly
application-dependent.
2.2 Heuristic Algorithms
Heuristic algorithms are used for the
static and dynamic tasks assignment problem.
Many of these algorithms apply only to the
special case where the tasks are independent
i.e. with no precedence constraints. Heuristic
scheduling algorithms are used in
heterogeneous computing environments. These
algorithms use historical data of execution
time and system load and explicit constraints
to schedule jobs.
2.3 Non-Evolutionary Random
Scheduling Algorithm
Non-evolutionary random scheduling
(RS) algorithm is used for efficient matching
and scheduling of inter-dependent tasks in a
distributed heterogeneous computing (DHC)
system. RS is a succession of randomized task
orderings and a heuristic mapping from task
order to schedule. Randomized task ordering is
effectively a topological sort where the
outcome may be any possible task order for
which the task precedent constraints are
maintained.
2.4 Fault Tolerant Dynamic
Scheduling Algorithm
Manimaran and Murthy (1997) [4]
proposed an algorithm for dynamically
scheduling arriving real-time tasks with
resource and primary-backup-based fault-
tolerant requirements in a multiprocessor
system. This algorithm can tolerate more than
one fault at a time and employs techniques
such as distance concept, flexible backup
overloading and resource reclaiming to
improve the guarantee ratio of the system.
They address the problem of building
a reliable and highly-available grid service by
replicating the service on two or more hosts
using the primary-backup approach. The
primary goal is to evaluate the ease and
efficiency with which this can be done, by first
designing a primary-backup protocol using
Open Grid Services Infrastructure (OSGI).
2.5 Primary-Backup Approach
Primary-backup approach, also called
passive replication strategy. In this approach a
backup is executed when its primary cannot
complete execution due to processor failure. It
does not require fault diagnosis and is
guaranteed to recover all affected tasks by
processor failure. Most works using the
primary-backup approach consider scheduling
of independent tasks.
2.5.1 Backup Overloading and
Overlapping
Backup overloading is used to reduce
replication cost of independent task which
allows scheduling backups of multiple
primaries on the same or overlapping time
interval on a processor.
In Backup Overlapping, for example,
two primary copies are scheduled on processor
1 and processor 3 and their backups are
scheduled in an overlapping manner on
processor 2.
2.6 Backup Schedules
After the earliest possible start time
for a backup on all processor is determined, the
time window that this backup can be scheduled
on all processor is determined which is
between this time and its deadline. Primary
schedules and non over loadable backup
schedules that are scheduled on the time
window can be identified. These backup
schedules could be scheduled for independent
tasks or dependent tasks as interleaving
technique is allowed.
3 PROPOSED WORK
The proposed system integrates
resource and data source from different
geographical environment. In this system,
location of resource and data source is
identified. There exist a fault-detection
mechanism such as fail-signal and acceptance

www.ijcat.com 776
test to detect processor and task failures. If a
failure is detected in the primary, the backup
will execute. Backup resources are designed
with replication factors. Impact of
communication protocols is considered.
Communication protocols are used to
distribute the message of each task to grid
resource.
3.1 Scheduling Strategies
The resources and data source are
managed from different environment. The
location of resources and data sources is
identified. There exist a fault-detection
mechanism such as fail-signal and acceptance
test to detect processor and task failures. If a
failure is detected in the primary, the backup
will execute. Backup resources are designed
with replication factors. Backup overloading is
used for scheduling backups of multiple
primaries on the same or overlapping time
interval on a processor. Resource reclaiming is
also invoked when the primary completes
earlier than its estimated execution time. It is
necessary so that the backup slot can be
released timely for new tasks.
MRC-ECT algorithm is used to
schedule the backup of independent job. MCT-
LRC algorithm is used to schedule the backup
of dependent job. For independent tasks,
scheduling of backups is independent and
backups can overload as long as their primaries
are schedule on different processors. Backup
scheduling and overloading of dependent tasks
are nontrivial and the constraint is that the
backup of second task can only start after
backup of first task finishes and must not be
schedule on the processor where primary of the
first task is located.
3.1.1 MRC-ECT Algorithm
MRC-ECT algorithm is used for
scheduling backup of independent tasks. The
objective is to improve resource utilization.
For all processor besides the one where the
primary is scheduled on, boundary schedules
within the time window are considered and
their replication cost is compared. This
algorithm first considers the left boundary
schedules of the time window. It is guaranteed
to find an optimal backup schedule in terms of
replication cost for a task.
3.1.2 MCT-LRC Algorithm
MCT-LRC algorithm is used for
scheduling backup of independent tasks. The
objective is to reduce job rejection. For all
processor besides the one where the primary is
scheduled on, boundary schedules within the
time window are considered and the boundary
schedule which can complete earliest is
chosen. This algorithm first considers the left
boundary schedules of the time window. Then,
all existing schedules within or overlapping
with the time window are examined one by
one. The algorithm calculates replication cost
of the earliest schedule on the current
processor and records it.
3.2 Communication Protocols
Different communication protocols
are used in grid environment. Transmission
Control Protocol (FTP) and User Datagram
Protocol (UDP) are used for data transmission.
Grid File Transfer Protocol (FTP) is used for
data transmission. It is used to transfer files in
parallel manner. These protocols are used to
distribute the message of each task to grid
resource. The system analyses the data
transmission in task failures.
4. CONCLUSION
In this paper, for Grid systems, we
addressed the problem of fault- tolerant
scheduling of jobs in heterogeneous
environment. We considered the impact of
communication protocols. Algorithms MRC-
ECT and MCT-LRC for independent and
dependent tasks respectively do not require
sampling. These algorithms can schedule
backups in a much faster way in heterogeneous
environment. .
REFERENCES
[1] Aikebaier.A, Makoto Takizawa,
Abawajy. J.H (2004), "Fault-Tolerant
Scheduling Policy for Grid Computing
Systems" Proceeding on Parallel and
Distributed Processing Symposium
(IPDPS)..
[2] Al-Omari, R., Somani, A.K., and
Manimaran, G. (2001),”A New Fault-
Tolerant Technique for improving
Schedulability in Multiprocessor Real-
Time Systems” Proceedings on
Parallel Distributed Processing
Symposium (IPDPS).

www.ijcat.com 777
[3] Foster .I and Kesselman .C
(2004),“Grid: Blueprint for a Future
Computing Infrastructure". Morgan
Kaufmann
[4] Subbiah A. and Blough D.
(2004),”Distributed Diagnosis in
Dynamic Fault Environments” Parallel
and Distributed Systems.
[5] Qin X. and Jiang H. (2006),”A Novel
Fault tolerant Scheduling Algorithm
for precedence constrained Tasks in
Real-Time Heterogeneous Systems”
Parallel Computing.
Author:
Mrs.R.Kohila received
M.E(CSE) degree from
Kongu Engineering
College(Affiliated to
Anna University,
Autonomous),
Perundurai, India in 2011 and MCA
degree from Bannari Amman Institute
of Technology (Affiliated to Anna
University), Sathyamangalam, India,
in 2009 and B.Sc., degree from Trinity
College of Arts and Science for
women (Affiliated to Periyar
University), Namakkal,India, in 2006.
She has the experience in Teaching of
3+Years. Now she is currently
working as an Assistant Professor in
V.S.B Engineering College, Karur,
Tamil Nadu, and India. His research
interests include Data Mining,
Advanced Data Bases, Computer
Networks etc. She had presented
papers in 2 National Conferences so
far.

Efficient Resource Management Mechanism with Fault Tolerant Model for Computational Grids

More Related Content

What's hot (16)

Viewers also liked (16)

Similar to Efficient Resource Management Mechanism with Fault Tolerant Model for Computational Grids (20)

More from Editor IJCATR (20)

Recently uploaded (20)

Efficient Resource Management Mechanism with Fault Tolerant Model for Computational Grids