Grenze International Journal of Engineering and Technology, Jan Issue

Comparative Study on Calculating CPU Burst Time

using Different Machine Learning Algorithms
Kuldeep Vayadande1, Naman Agarwal2, Srushti Shevate3, Chinmayee Sawakare4, Piyush Parakh 5 and
Snehankit Khandare6
1-6
Vishwakarma Institute of Technology/Department of Artificial Intelligence and Data Science, Pune, India
Email: [email protected], {snehankit.khandare20, srushti.shevate20, chinmayee.sawakare20, piyush.parakh20,
naman.agarwal20}@vit.edu

Abstract—This paper gives a thorough analysis of works on the subject of expert assessment of
CPU burst time using various machine learning algorithms. Knowing how long the CPU bursts
for the processes will last is necessary for some CPU scheduling algorithms like SJF and SRTF
to function. In particular, the non-preemptive SJF scheduling algorithm estimates the process
that will be performed by the CPU in the least amount of burst time. One effective way of
predicting CPU burst duration is an ML-based algorithm that estimates the burst-time of the
processes. Throughout the study, we discovered that the effectiveness of different machine-
learning approaches relies on the applications to which they are put. Our examination of the
literature not only argues that these methods are competitive with conventional estimators on a
single data set, but also demonstrates that they are responsive to the training data.

Index Terms— CPU burst time, scheduling algorithms, SJF, ML algorithms.

I. INTRODUCTION
The mechanism for granting access to tasks, processes, threads, or data flow is called a scheduling algorithm.
The requirement to execute numerous processes simultaneously has prompted the creation of scheduling
algorithms in today's multiprogramming and modern systems. There are several scheduling ways suggested to
decide which tasks should be executed on priority basis. Some algorithms are preferred over others including
Round Robin, SJF, and FCFS. This comparison's goal is to evaluate whether a method is better suited to certain
tasks in the ready queue. Various algorithms, like FCFS, Shortest Job First, and others, which depend on time
frame parameters like Arrival Time, Burst Time, Waiting Time, etc., are used to execute numerous processes in
an optimised manner. The magnitude of the CPU burst must be known in order to employ SJF scheduling or
SRTF. Burst time is how long a process requires to execute, expressed in milliseconds. The burst time accounts
for the CPU time used by a process. The time spent doing I/O is not taken into account. SJF and SRTF
algorithms require processes which have short duration bursts over those with longer ones to make sure that the
smaller process will finish and exit the system faster than the bigger one. It is not simple to determine the next
CPU burst size for a process; it requires additional work and computations.
Earlier, the length of CPU bursts was measured using mathematical formulas. That traditional way is called
“Exponential Averaging (EA)”, this technique uses the previous history of CPU bursts to predict the next
burst. When multi-level platforms from various places in computational grids are employed to deliver a variety
of services to a shared objective; work scheduling is seen as a crucial element in the infrastructure of

Grenze ID: 01.GIJET.9.1.326

© Grenze Scientific Society, 2023
a computational grid. Grid scheduling is the process of allocating resources across several administrative
domains to Grid jobs. Our research concentrates on the many ML approaches that are employed and can be used
for efficient estimation of CPU-burst time.

II. LITERATURE SURVEY

A Pourali, Amir Rahmani [1] A variety of time-stamped data, in particular CPU-burst data series, may be
forecasted with a certain amount of accuracy by using expert machines like fuzzy systems. Some perks of fuzzy
logic involve improved adaptability and optimal speed. The significance of the suggested technique is its
capacity to anticipate a process's forthcoming CPU burst time based on previous information.
Tarek H, Sadam Al-Azani, Omar Obaidellah [2] By utilising ML techniques such as SVM, DT, ANN, and K-
NN, the report recommends a technique to evaluate the CPU burst duration for activities. The "GWA-T-4 Auver
Grid" grid workload was implemented in the suggested method to apply the attribute feature selection
approaches choose the most solution that will work. depending on the outcomes of the experiment. Despite a
strong link between the business factors and the burst Computational time, K-Nearest Neighbour consistently
outperforms further Machine Learning algorithms on the grounds of RAE & CC.
Putera, Andysah, Utama, Siahaan [3] Calculations using three algorithms show the varying average waiting
times. For a short while, first come first served is preferable. Shortest Job First is preferable when multiple
processes are entering the CPU simultaneously. The best way to change the needed average waiting time is to
use the Round Robin approach. Greater SJF or FCFS values will be established at round robin response time. All
algorithms work well, although the rate of process advancement depends on the CPU load.
Sirisha B, Suresh K, Diya D S, Soumya [4] The suggested remedy employs a number of approaches, including
k-nearest neighbours, decision learning, and linear regression, to determine the Central Processing Unit burst
length for the method in the ready list. This strategy is more effective than earlier CPU scheduling techniques
like SJF and SRTF.
J R Indusree, B Prabadevi [5] The experiment results clearly show that the proposed algorithm offers improved
performance. In comparison to the RR algorithm, there are a lot fewer context switch. Additionally, it shortens
turnaround and average waiting times.
Bushra Hamid, Inayat Ur-Rehman [6] This research suggests that the running performance of diverse scheduling
algorithms with the recently proposed scheduling approach. In comparison to round robin and pre-emptive SJF,
the proposed approach has less context switching, according to the comparison results. Application of the
suggested method can effectively maximise system performance by reducing context switching.
Mahesh Kumar M.R, Renuka Rajendra.B, Niranjan C.K, Sreenath .M [7] The operation that has a shorter burst
time than the preceding ones is chosen from the ready queue by a conventional SJF scheduling algorithm. This
can be resumed as long as the final operation in the waiting list is completed. This technique always requires the
CPU to execute the next operation with their burst length. This information isn't always accessible, though.
Dr Vandana [8] This study aims to update the EAF (Exponential Average Formula) utilising fuzzy systems. Prior
to a process's execution, the suggested method's major goal is to predict and estimate its execution time. In order
to obtain high flexibility and the desired calculation speed, fuzzy inference engines are used.
Ms. Vaibhavi Pandya, Mr. Indra Jeet Rajput [9] For the SRR scheduling, the execution's time quantum is
determined beforehand at a specific time, however in this suggested scheduling. Although time quantum's
algorithm is not predefined, it is based on freshly established work sets, modified. As a result, the approach does
not employ conventional technique, so is more efficient than a static round robin.
Parul Prakram, Mohit Prakram [10] SJF is among the top CPU scheduling algorithms. Despite this, the CPU
scheduling algorithm cannot be used in real world due to the difficulty in determining the precise time spent
during a process's execution before it has truly executed. Consequently, utilising the suggested Fuzzy based
utilising an SJF scheduling technique that has to be aware of approaching CPU-burst times for processes. It will
be realised in a working computer system.

2864
III. HELPFUL HINTS

A. Figures and Tables

TABLE I. COMPARISON TABLE OF ALL PAPERS

SR
ALGORITHM
.N TITLE JOURNAL NAME YEAR DESCRIPTION
USED
O
It has been suggested how to figure out how long
CPU Burst International Journal
a CPU burst should last for each task. It uses a
Time using of Advanced K-Nearest
variety of approaches, including k-nearest
Machine Research in Science, Neighbour, Decision
1. 2022 neighbour, decision learning, and linear
Learning Communication and learning, and Linear
regression. This strategy is more effective than
Algorithm’s Technology Regression.
earlier CPU scheduling techniques like SJF and
[3] (IJARSCT) [3]
SRTF.

A Machine
Learning- In this research, they have proposed a method for
Based predicting the duration of CPU burst using
3rd International
Approach to Machine Learning methods. Additionally, in
Conference on
Estimate the Feature Selection order to choose the most crucial properties, the
Artificial
2. CPU-Burst 2015 CPU Scheduling "GWA-T-4 Auver Grid" grid burden was put
Intelligence,
Time for Algorithm through attribute feature selection procedures.
Modeling and
Processes in The best Machine Learning method is K-NN in
Simulation [2]
the practically every case when concerned with
Computation associated coefficient and Error.
al Grids [2]

Prediction of
Length of the In this situation, the standard Shortest Job First
2nd International
Next CPU selects the operation from the waiting list where
Conference on Conventional SJF
Burst in SJF the CPU burst duration is minimum. This
Current Trends in Scheduling
3. Scheduling 2014 strategy always needs the CPU to run the next
Engineering and algorithm, Machine
Algorithm operation with their execution times until
Technology, Learning
using Dual achieving the finish we practically can’t estimate
ICCTET’14 [7]
Simplex the upcoming CPU burst time.
Method [7]

A Fuzzy-
A variety of time-stamped data, in particular
based
CPU-burst data series, may be forecasted with a
Scheduling
International certain amount of accuracy by using expert
Algorithm for
Association of machines like fuzzy systems. Some perks of
prediction of Fuzzy systems,
Computer Science fuzzy logic involve improved adaptability and
4. next CPU- 2009 Shortest Process
and Information optimal speed. The significance of the suggested
burst time to Next (SPN)
Technology - Spring technique is its capacity to anticipate a process's
implement
Conference [1] forthcoming CPU burst time based on previous
Shortest
information.
Process Next
[1]

A Fuzzy Built
Methodology
towards A lower bound "a," an upper bound "b," and a
Predicting the parameter "m" are used to create a triangular
CPU Scheduling
Next CPU- THINK INDIA linear model, which is what the method in this
5. 2019 algorithms, SJF
Burst for SJF JOURNAL [10] case does. For the same objective, they used the
algorithm,
Scheduling Fuzzy constructed algorithm in order to calculate
Algorithm the exponential average Algorithm.
[10]

2865
 They used a set of processes that were randomly
Revised Journal of Average chosen, and they computed the results to forecast
Formula for International Exponential the CPU. Using MATLAB and C++, the updated
6. Estimating Academy of 2018 Formula (AEF), exponential average formula was compared to the
CPU-Burst Physical Sciences Fuzzy Inference original EAF. Additionally, they attempted to
[8] [8] System (FIS) present us with graphic comparisons using three
distinct numbers for α, such as 0, 1, and 12.

Enhanced
round robin The experiment's findings unmistakably
CPU IOP Conference demonstrate that the proposed method provides
FCFS algorithm,
scheduling Series: Materials better performance. In comparison to the RR
7. 2017 Priority scheduling
with burst Science and algorithm, there are a lot fewer context switch.
algorithm
time-based Engineering [5] Additionally, it shortens turnaround and average
time quantum waiting times.
[5]

Performance
Improvisation In terms of execution, the adaptive round robin
International Journal Dynamic time,
in Scheduling scheduling is an effective scheduling technique.
for Technological quantum time
8. through 2017 The suggested system's primary emphasis is on
Research in quantum context
Optimal creating task sets and subsequently use the
Engineering [9] switching
Resource algorithm for adaptive round robin
Utilization [9]

Here are the outcomes of three methods that

serve as illustrations of the varied mean waiting
Comparison times. For small periods, the FCFS functions
Analysis of effectively. The SJF is better if the
CPU operations train and develop the CPU
IJEDR | Volume 4, SJF, Round Robin
9. Scheduling: 2016 simultaneously. For altering the needed mean
Issue 3 [4] and CPU scheduling
FCFS, SJF waiting times, the RR technique works well.
and Round RR quantum frequency will set a greater value
Robin [4] for SJF or FCFS. All algorithms operate as
intended, but CPU load controls how fast a
process proceeds.

The hybrid scheduling algorithm was created by

An Optimized them. The shortest job first technique plus the
Shortest job constraint on residual burst length of the
first Journal of Applied executing process method make up this
Hybrid Scheduling
Scheduling Environmental and algorithm. The rationale behind such a strategy is
10. 2015 algorithm,
Algorithm for Biological Sciences that it starts this procedure on the task with the
Preemptive SJF
CPU [6] shortest burst time at launch. They also noted that
Scheduling the scheduling technique is proactive. As a result,
[6] when a new job joins, it can decide if it wants to
preempt an active process.

2866
B. Comparison Graphs

Figure 1. Analogy of actual burst time and predicted burst time using Figure 3. Accuracy Rate for ML Algorithms
ML algorithms

Figure 2. Comparison of Coefficient Comparison and Relative Figure 4. Plotting Burst time and Arrival Time using Scatter Plot
Absolute Error

Figure 5. Round Robin Quantum Time Calculation

IV. SCOPE OF IMPLEMENTATION

In this paper, further advanced and improved algorithms and building models to have a detailed comparative
analysis of the CPU burst time dataset can be discussed. More features can be included for comparing algorithms
such as accuracy, F1 score etc. In further research, more advanced ML models to predict the CPU burst time can
be .

V. CONCLUSION
Discussing the different Machine-Learning Techniques utilised for CPU Burst time estimate is the major
contribution of this review. As a conclusion it was found that in order to determine the optimal method to
calculate CPU burst time for processes, a variety of techniques including linear regression, logistic regression,

2867
decision trees, and KNN should be used. We should employ the approach that best fits the model while keeping
in mind its constraints and the primary goal of enhancing its performance and effectiveness.
Our research also supports the idea that there is no single machine learning method that is ideal. Because of this,
a greater understanding of the applicability and scope of a number of the strategies mentioned is urgently
needed.

REFERENCES
[1] Pourali, Abdolghader, and Amir Masoud Rahmani. "A Fuzzy-based Scheduling Algorithm for prediction of next CPU-
burst time to implement Shortest Process Next," International Association of Computer Science and Information
Technology-Spring Conference. IEEE, 2009.
[2] Helmy, Tarek, Sadam Al-Azani, and Omar Bin-Obaidellah. "A machine learning-based approach to estimate the CPU-
burst time for processes in the computational grids" 3rd International Conference on Artificial Intelligence, Modelling
and Simulation (AIMS). IEEE, 2015.
[3] Samal, Prathamesh, Sagar Jha, and Raman Kumar Goyal. "CPU Burst-Time Estimation using Machine Learning," IEEE
Delhi Section Conference (DELCON). IEEE, 2022.
[4] Siahaan, Andysah Putera Utama. "Comparison analysis of CPU scheduling: FCFS, SJF and Round Robin," International
Journal of Engineering Development and Research 4.3 (2016): 124-132.
[5] Indusree, J. R., and B. Prabadevi. "Enhanced round robin CPU scheduling with burst time based time quantum," IOP
Conference Series: Materials Science and Engineering. Vol. 263. No. 4. IOP Publishing, 2017.
[6] Hamayun, Maryam, and Hira Khurshid. "An optimized shortest job first scheduling algorithm for CPU scheduling," J.
Appl. Environ. Biol. Sci 5.12 (2015): 42-46.
[7] Kumar, MR Mahesh, et al. "Prediction of length of the next CPU burst in SJF scheduling algorithm using dual simplex
method," Second International Conference on Current Trends in Engineering and Technology-ICCTET 2014. IEEE,
2014.
[8] Vandana "Revised Formula for Estimating CPU-Burst,” Journal of International Academy of Physical Sciences Vol. 22
No. 4 (2018)
[9] Ms.Vaibhavi Pandya, Mr.Indr Jeet Rajput, “PERFORMANCE IMPROVISATION IN SCHEDULING THROUGH
OPTIMAL RESORCE UTILIZATION,” International Journal For Technological Research In Engineering Volume 4,
Issue 11, July-2017
[10] Prakram, Parul, and Mohit Prakram. "A Fuzzy Built Methodology towards Predicting the Next CPU-Burst For SJF
Scheduling Algorithm," Think India Journal 22.17 (2019): 3851-3859.

2868