Project contexts switching and the probability of overhead

Project Summary

This paper is about introducing a new scheduling
technique in a real-time system which is
defined based on the paper as the computation not only depends on producing a
correct output, but the output is also delivered within a deadline. The
identified problem for proposing such technique is because both EDF and LLF
have many contexts switching and the probability
of overhead occurrence causing degraded performance of the system. The proposed technique is a dynamic
scheduling algorithm that combines both
the EDF and LSF algorithms to reduce the problems that had been identified
before.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

 

Background

Earliest Deadline First (EDF) work by giving
priorities based on the smallest deadline where it will then have the highest
probability. This scheduling algorithm was introduced by Liu and Layland in
1973 while Least Laxity First (LLF) and Maximum Urgency First (MUF) was
introduced by David B. Stewart and Pradeep K. Khosla where for LLF work by
selecting task with the minimum laxity to execute and MUF is a combination of
fixed and dynamic priority scheduling where each task is given an urgency. All
the algorithms have its own problems based on the finding of this paper. The
problems have then initiated the idea to
resolve the problems of the discussed algorithms. However, the proposed work is
only focused on comparing itself with only two dynamic algorithms which are the EDF and LLF. The identified problems
for EDF are high average waiting time, many contexts switching, does not let
tasks be executed with their minimal time and leads to the occurrence of overhead. The LLF on the other
hand facing the same problem with the EDF which has many context switches
occurred when two processes have similar laxities. This paper is a
collaboration with different universities from a different country which is from the country of Ethiopia and India.
Three of the authors are from Ethiopia and only one of them are from India. The
collaborated universities are between the Debre Markos University, Adama
Science and Technology University and St Mary’s College of Engineering and
Technology from India.

Process

The process of making this research a success is by
undergoing a simulation of the proposed algorithm with the previously developed
dynamic scheduling algorithms. It was simulated based on the parameters of
arrival time, execution, and relative
deadline while the schedulability of the proposed system was compared with the
schedulability of EDF where it considered many metrics like average waiting
time of the task, context switching,
system overhead and optimality. The three main parameters on the dataset
preparation of this paper have it own
rules which the first one is for the arrival time where each task is computed
using uniformly distributed random
integer number to make the computation easier. The range taken was from 1 to 30
as in real case more than one task will arrive at the same time and by doing
so, at least 20% of the total dataset will be ready at the same time. The second
rule is for the execution time where each task is computed using the same
technique with the arrival time rule but only ranging from 1 to 6 based on the
assumption of each task should have the maximum execution time of 6 units. It
is also stated in the rule that the execution time should be less than (deadline – arrival time) of each task. The third rule applies for the deadline where each task is determined by the logic that its deadline should be greater than
(arrival time + execution time) and
vice versa. Figure 1 is the proposed algorithm of this paper. It works by
determining the precise slack time and a deadline
of tasks where it will then give priority to the task and this algorithm is a
dynamic priority scheduling algorithm. It basically
works with the same concept of EDF except that it will also consider the
slack time of the tasks and let it continue to run the current task so that one
will not miss its deadline.

From the Table
1, have three algorithms has been used to test using same sample data. Firstly
by used scheduling with EDF, this paper found that sample data can be
scheduled, and the number of context switching and average waiting time can be
counted clearly. The second LLF
scheduling has been used and found that the result as same as EDF scheduler.
Lastly by used proposed algorithms scheduling, the result shows that the previous
proposed dynamic algorithms are less efficient
in term of context switching and average waiting time, which bottlenecks for
the system has overall performance.

Based on the finding in Table 2 and Figure 2, finding a show
that the number of context switching is high and very dynamic due to the Least
Slack First algorithms. But different to the proposed algorithms, it shows the number context switching is low and
mostly less than EDF and LST scheduler.

In Table 3 and Figure 3, if the number of context
switching is high than the average waiting time also become high except
switching is due to pre-empting the task have high
priority by the task having low slack time. In the finding, it shows that the
proposed algorithms have a low average waiting time than both EDF and LST
scheduling techniques.     

In this finding also has mentioned on overhead performance, due to experiment that has been
done, it can say that proposed algorithms
have a low overhead than both EDF and
LST. In the proposed system the probability of tie to occur is (LST n EDF) where LST is laxity tie and EDF is deadline
tie, both want to execute at the same time.LST intersection EDF this is
because the proposed scheduler takes both slack and deadline into
consideration, so the scheduled will be confused to execute the task. This will show that, the
number of elements in (LST n EDF) less and equal to the element in individual
sets. Therefore, the proposed system is better in term of occurrence of
overhead. Based on all of the finding,
it can be concluded and prove that the real-time
dynamics scheduling algorithms are the
best and suit to the uni-processor system.

 

Recommendations for further
action

 

In future work to make
the best real-time
dynamics scheduling algorithms performance, this
can be recommended to combined some logic data such as slack and deadline into
multiprocessor and should use more logical and reasonable data.

x

Hi!
I'm Rita!

Would you like to get a custom essay? How about receiving a customized one?

Check it out