Task Migration with EDF-RM Scheduling Algorithms in Distributed System
Rashmi Sharma
Department of CSE and ICT
Jaypee University of Information Technology,
Waknaghat, Solan-173234, Himachal Pradesh, India
rashmi.nov30@gmail.com
Nitin
Department of CSE and ICT
Jaypee University of Information Technology,
Waknaghat, Solan-173234, Himachal Pradesh, India
delnitin@ieee.org
Abstract—Earliest deadline first and Rate Monotonic are the two
well-known scheduling algorithms to meet the deadlines for task
in RTS. For single processor, these tasks work well but for more
than one node, overloading in utilization usually creates problem.
Task migration and duplication are those two concepts that
balance the load of distributed system due to which overloading
problem has been resolved. This paper explains the processing of
task migration with EDF-RM (simultaneously) scheduling
algorithm by using Global Scheduling.
Keywords- Real Time Distributed System (RTDS); Task Migration;
Earliest Deadline First (EDF); Rate Monotonic (RM).
I. INTRODUCTION AND MOTIVATION
Management of real time tasks is somewhat uncertain in
distributed system. In real time uniprocessor system after
crossing the limit of utilization tasks start missing their
deadline. However, in RTDS, mostly tasks distribute to other
processors of the system and parallel execution takes place.
Therefore, the rejection is overruled. The balancing of
distributed tasks sustain by task migration or duplication
concept. For the maintenance of proper resource utilization,
task migration and duplication are the favorable solutions.
The system under consideration is a Real Time Distributed
System (RTDS), which is by definition “A Distributed System
is having Real Time Properties”. The architecture of processors
in distributed system can be homogenous as well as
heterogeneous. In homogenous system processors share the
similar architecture and dissimilar in heterogeneous. Here we
are using homogenous system [1].
In Real time system, every task has a deadline (by that time
task should execute). For scheduling of real time tasks, Rate
Monotonic and Earliest deadline first are two well-known
scheduling algorithms under which execution of tasks based on
its period of arrival or deadline as well. RM algorithm works
with Static Priority Scheduling (offline tasks) and EDF
algorithm is with Dynamic Priority Scheduling (online tasks).
The arrival of tasks in a particular system can be periodic,
aperiodic or sporadic. Mostly systems set aside the arrival of
tasks periodically because the period of task arrival is fixed and
these tasks are able to meet their respective deadline (relative
deadline) [2].
Every scheduling algorithm has its own merits and
demerits, like EDF assign priority based on deadline of the task
and working well for single processor in underloading
condition but working inefficiently in overloaded case [3]. RM
is a static algorithm and priorities based on periods but it is not
as capable as dynamic algorithms for underloaded conditions
[4] but performing well in overloaded as compared to Dynamic
algorithms. In Distributed System, this overloaded and
underloaded problem has been balanced by using the task
migration method.
For scheduling of periodic task systems on Distributed
System there have been two approaches: partitioning and
global scheduling. In global scheduling, single priority ordered
queue of eligible tasks have maintained and scheduler selects
the highest priority task for execution. However, in partitioning
scheduling, each task is assigned to single processor and
processors are scheduled independently [5]. Out of these two
schedulers, previous one requires task migration and in last task
migration is prohibited [6].
Load balancing (in RTDS) is handled by using task
migration with optimal scheduler and for the execution of real
time tasks either RM or EDF algorithms are used according to
the demand of the system. This paper explains RM and EDF
together with migration.
The rest of the paper organizes as follows: Section II
explains the real time scheduling algorithms (RM and EDF)
and migration in distributed system, Section III describes the
proposed algorithm of RM and EDF (with Migration) with
examples. Further Conclusion and future work will be
discussed.
II. PRELIMINARIES AND BACKGROUND
The major approach of scheduling multiprocessors real time
tasks is partitioned scheduler in which only one task can
execute at a time on particular processor; a bin-packing
problem is presented here in the same. On the other side, global
scheduling is another option of partitioned scheduling which
maintains a global task queue from which tasks are sent out to
any vacant processor according to a global priority scheme [7].
In terms of task migration, we can say that in partitioned
scheduler it is not permitted whereas in global scheduling
migration can be done. Load balancing in distributed system is
necessary which is achieved by the use of task migration
concept. Hence, in our system we are using global scheduler
for scheduling the tasks of homogenous RTDS.
The arrival of real time tasks is periodic in nature because
the execution of one task on or before the arrival of next task is
necessary otherwise, there will be a chance of missing the
deadline. Real time periodic tasks have following parameters:
2012 International Conference on Advances in Computing and Communications
978-0-7695-4723-7/12 $26.00 © 2012 IEEE
DOI 10.1109/ICACC.2012.42
182
2012 International Conference on Advances in Computing and Communications
978-0-7695-4723-7/12 $26.00 © 2012 IEEE
DOI 10.1109/ICACC.2012.42
182