Machine Learning for

Cloud Management
Machine Learning for
Cloud Management

Jitendra Kumar
Ashutosh Kumar Singh
Anand Mohan
Rajkumar Buyya
First edition published 2022
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742

and by CRC Press

2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

© 2022 Jitendra Kumar, Ashutosh Kumar Singh, Anand Mohan, Rajkumar Buyya

CRC Press is an imprint of Taylor & Francis Group, LLC

Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot as-
sume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have
attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders
if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please
write and let us know so we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or
utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including pho-
tocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission
from the publishers.

For permission to photocopy or use material electronically from this work, access www.copyright.com or contact the
Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that are
not available on CCC please contact [email protected]

Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used only for iden-
tification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Names: Kumar, Jitendra, 1975- author. | Singh, Ashutosh Kumar, author. |

Mohan, Anand (Of Indian Institute of Technology), author. | Buyya,
Rajkumar, 1970- author.
Title: Machine learning for cloud management / Jitendra Kumar, Ashutosh
Kumar Singh, Anand Mohan, Rajkumar Buyya.
Description: First edition. | Boca Raton : CRC Press, 2022. | Includes
bibliographical references and index.
Identifiers: LCCN 2021027713 | ISBN 9780367626488 (hardback) | ISBN
9780367622565 (paperback) | ISBN 9781003110101 (ebook)
Subjects: LCSH: Cloud computing. | Machine learning.
Classification: LCC QA76.585 .K85 2022 | DDC 004.67/82--dc23
LC record available at https://lccn.loc.gov/2021027713

ISBN: 978-0-367-62648-8 (hbk)

ISBN: 978-0-367-62256-5 (pbk)
ISBN: 978-1-003-11010-1 (ebk)

DOI: 10.1201/9781003110101

Publisher’s note: This book has been prepared from camera-ready copy provided by the authors.
 Dedicated to,
My wife: Gita, daughter: Aru, and Parents

∼Jitendra Kumar

Anushka, Aakash, Akankshya, and Parents

∼Ashutosh Kumar Singh

My wife: Sudha Mohan, son: Ashish Mohan, daughter: Amrita

Mohan, and Late parents

∼Anand Mohan

My international collaborators and team members in Melbourne

CLOUDS Lab!

∼Rajkumar Buyya
Contents

List of Figures xi

List of Tables xvii

Preface xix

Author xxi

Abbreviations xxv

Chapter 1  Introduction 1

1.1 CLOUD COMPUTING 1

1.2 CLOUD MANAGEMENT 2
1.2.1 Workload Forecasting 3
1.2.2 Load Balancing 4
1.3 MACHINE LEARNING 5
1.3.1 Artificial Neural Network 5
1.3.2 Metaheuristic Optimization Algorithms 6
1.3.3 Time Series Analysis 7
1.4 WORKLOAD TRACES 7
1.5 EXPERIMENTAL SETUP & EVALUATION METRICS 8
1.6 STATISTICAL TESTS 9
1.6.1 Wilcoxon Signed-Rank Test 10
1.6.2 Friedman Test 10
1.6.3 Finner Test 10

Chapter 2  Time Series Models 13

2.1 AUTOREGRESSION 14
2.2 MOVING AVERAGE 14
2.3 AUTOREGRESSIVE MOVING AVERAGE 15

vii
viii  Contents

2.4 AUTOREGRESSIVE INTEGRATED MOVING AVERAGE 15

2.5 EXPONENTIAL SMOOTHING 17
2.6 EXPERIMENTAL ANALYSIS 17
2.6.1 Forecast Evaluation 17
2.6.2 Statistical Analysis 21

Chapter 3  Error Preventive Time Series Models 25

3.1 ERROR PREVENTION SCHEME 25

3.2 PREDICTIONS IN ERROR RANGE 27
3.3 MAGNITUDE OF PREDICTIONS 28
3.4 ERROR PREVENTIVE TIME SERIES MODELS 29
3.4.1 Error Preventive Autoregressive Moving Average 29
3.4.2 Error Preventive Autoregressive Integrated Moving Average 35
3.4.3 Error Preventive Exponential Smoothing 38
3.5 PERFORMANCE EVALUATION 47
3.5.1 Comparative Analysis 47
3.5.2 Statistical Analysis 53

Chapter 4  Metaheuristic Optimization Algorithms 59

4.1 SWARM INTELLIGENCE ALGORITHMS IN PREDICTIVE MODEL 59

4.1.1 Particle Swarm Optimization 60
4.1.2 Firefly Search Algorithm 61
4.2 EVOLUTIONARY ALGORITHMS IN PREDICTIVE MODEL 62
4.2.1 Genetic Algorithm 63
4.2.2 Differential Evolution 63
4.3 NATURE INSPIRED ALGORITHMS IN PREDICTIVE MODEL 64
4.3.1 Harmony Search 64
4.3.2 Teaching Learning Based Optimization 66
4.4 PHYSICS INSPIRED ALGORITHMS IN PREDICTIVE MODEL 67
4.4.1 Gravitational Search Algorithm 68
4.4.2 Blackhole Algorithm 69
4.5 STATISTICAL PERFORMANCE ASSESSMENT 71
 Contents  ix

Chapter 5  Evolutionary Neural Networks 75

5.1 NEURAL NETWORK PREDICTION FRAMEWORK DESIGN 75

5.2 NETWORK LEARNING 77
5.3 RECOMBINATION OPERATOR STRATEGY LEARNING 78
5.3.1 Mutation Operator 78
5.3.1.1 DE/current to best/1 78
5.3.1.2 DE/best/1 78
5.3.1.3 DE/rand/1 78
5.3.2 Crossover Operator 79
5.3.2.1 Ring Crossover 79
5.3.2.2 Heuristic Crossover 79
5.3.2.3 Uniform Crossover 80
5.3.3 Operator Learning Process 80
5.4 ALGORITHMS AND ANALYSIS 83
5.5 FORECAST ASSESSMENT 86
5.5.1 Short Term Forecast 87
5.5.2 Long Term Forecast 88
5.6 COMPARATIVE ANALYSIS 92

Chapter 6  Self Directed Learning 97

6.1 NON-DIRECTED LEARNING-BASED FRAMEWORK 97

6.1.1 Non-Directed Learning 98
6.2 SELF-DIRECTED LEARNING-BASED FRAMEWORK 99
6.2.1 Self-Directed Learning 100
6.2.2 Cluster-Based Learning 100
6.2.3 Complexity analysis 102
6.3 FORECAST ASSESSMENT 103
6.3.1 Short Term Forecast 103
6.3.1.1 Web Server Workloads 104
6.3.1.2 Cloud Workloads 104
6.4 LONG TERM FORECAST 106
6.4.0.1 Web Server Workloads 106
6.4.0.2 Cloud Workloads 107
6.5 COMPARATIVE & STATISTICAL ANALYSIS 108
x  Contents

Chapter 7  Ensemble Learning 121

7.1 EXTREME LEARNING MACHINE 121

7.2 WORKLOAD DECOMPOSITION PREDICTIVE FRAMEWORK 122
7.2.1 Framework Design 122
7.3 ELM ENSEMBLE PREDICTIVE FRAMEWORK 125
7.3.1 Ensemble Learning 126
7.3.2 Expert Architecture Learning 127
7.3.3 Expert Weight Allocation 129
7.4 SHORT TERM FORECAST EVALUATION 130
7.5 LONG TERM FORECAST EVALUATION 133
7.6 COMPARATIVE ANALYSIS 137

Chapter 8  Load Balancing 141

8.1 MULTI-OBJECTIVE OPTIMIZATION 141

8.2 RESOURCE-EFFICIENT LOAD BALANCING FRAMEWORK 142
8.3 SECURE AND ENERGY-AWARE LOAD BALANCING FRAMEWORK 146
8.3.1 Side-Channel Attacks 147
8.3.2 Ternary Objective VM Placement 148
8.4 SIMULATION SETUP 151
8.5 HOMOGENEOUS VM PLACEMENT ANALYSIS 151
8.6 HETEROGENEOUS VM PLACEMENT ANALYSIS 152

Chapter 9  Summary 155

Bibliography 159

Index 171
List of Figures

1.1 Service model view of cloud computing 2

1.2 Cloud resource management view 2
1.3 Cloud resource management approaches 3
1.4 Schematic representation of workload forecasting 4
1.5 Load balancing 5
1.6 Artificial neural network 6
1.7 An arbitrary optimization function with multiple local optima 7

2.1 Autoregression process 14

2.2 Moving average process 15
2.3 Autoregressive moving average process 15
2.4 Autoregressive integrated moving average process 16
2.5 Exponential smoothing process 17
2.6 Autoregression forecast results on MAE 18
2.7 Autoregression forecast results on MASE 18
2.8 Moving average forecast results on MAE 19
2.9 Moving average forecast results on MASE 19
2.10 Autoregressive moving average forecast results on MAE 20
2.11 Autoregressive moving average forecast results on MASE 20
2.12 Autoregressive integrated moving average forecast results on MAE 21
2.13 Autoregressive integrated moving average forecast results on MASE 21
2.14 Exponential smoothing forecast results on MAE 22
2.15 Exponential smoothing forecast results on MASE 22

3.1 Error preventive workload forecasting model 26

3.2 An example of error preventive and non-error preventive forecast 27
3.3 Prediction error range and segments 28
3.4 Error preventive ARMA forecast analysis for 10-minute prediction
interval 30

xi
xii  LIST OF FIGURES

3.5 Error preventive ARMA forecast analysis for 20-minute prediction

interval 31
3.6 Error preventive ARMA forecast analysis for 30-minute prediction
interval 32
3.7 Error preventive ARMA forecast analysis for 60-minute prediction
interval 33
3.8 R-score comparison of non-error preventive and error preventive ARMA 34
3.9 SEI comparison of non-error preventive and error preventive ARMA 35
3.10 MPE comparison of non-error preventive and error preventive ARMA 35
3.11 PER comparison of non-error preventive and error preventive ARMA 36
3.12 Positive magnitude comparison of non-error preventive and error
preventive ARMA 37
3.13 Negative magnitude comparison of non-error preventive and error
preventive ARMA 37
3.14 Error preventive ARIMA forecast analysis for 10-minute prediction
interval 39
3.15 Error preventive ARIMA forecast analysis for 20-minute prediction
interval 40
3.16 Error preventive ARIMA forecast analysis for 30-minute prediction
interval 41
3.17 Error preventive ARIMA forecast analysis for 60-minute prediction
interval 42
3.18 R-score comparison of non-error preventive and error preventive ARIMA 43
3.19 SEI comparison of non-error preventive and error preventive ARIMA 43
3.20 MPE comparison of non-error preventive and error preventive ARIMA 44
3.21 Positive magnitude comparison of non-error preventive and error
preventive ARIMA 45
3.22 Negative magnitude comparison of non-error preventive and error
preventive ARIMA 45
3.23 PER comparison of non-error preventive and error preventive ARIMA 46
3.24 Error preventive ES forecast analysis for 10-minute prediction interval 47
3.25 Error preventive ES forecast analysis for 20-minute prediction interval 48
3.26 Error preventive ES forecast analysis for 30-minute prediction interval 49
3.27 Error preventive ES forecast analysis for 60-minute prediction interval 50
3.28 R-score comparison of non-error preventive and error preventive ES 51
3.29 SEI comparison of non-error preventive and error preventive ES 51
3.30 MPE comparison of non-error preventive and error preventive ES 52
3.31 Positive magnitude comparison of non-error preventive and error
preventive ES 52
 LIST OF FIGURES  xiii

3.32 Negative magnitude comparison of non-error preventive and error

preventive ES 53
3.33 PER comparison of non-error preventive and error preventive ES 54
3.34 R-score performance relative improvement 55
3.35 SEI performance relative improvement 55
3.36 MPE performance relative improvement 56
3.37 Friedman test ranks of non-error preventive and error preventive models 56

4.1 Population-based metaheuristic optimization algorithms’ taxonomy 59

4.2 Forecast accuracy comparison of swarm intelligence based prediction
models 62
4.3 Wilcoxon test statistics of swarm intelligence based prediction models 62
4.4 Forecast accuracy comparison of evolutionary algorithms based pre-
diction models 65
4.5 Wilcoxon test statistics of evolutionary algorithms based prediction
models 65
4.6 Forecast accuracy comparison of nature-inspired algorithms based
prediction models 67
4.7 Wilcoxon test statistics of nature-inspired algorithms based prediction
models 68
4.8 Forecast accuracy comparison of physics-inspired algorithms based
prediction models 70
4.9 Wilcoxon test statistics of physics-inspired algorithms based prediction
models 71
4.10 Friedman test ranks of metaheuristic algorithms based prediction
models 71

5.1 Neural network-based workload prediction model 76

5.2 Ring crossover 79
5.3 Heuristic crossover 80
5.4 Uniform crossover 81
5.5 Learning period effect on forecast accuracy of self-adaptive differential
evolution algorithm based workload prediction model 86
5.6 Short term forecast assessment of self-adaptive differential evolution
algorithm based prediction model on 1-minute prediction interval 87
5.7 Short term forecast residuals of self-adaptive differential evolution
algorithm based prediction model on 1-minute prediction interval 88
5.8 Short term forecast assessment of biphase adaptive differential evolu-
tion algorithm based prediction model on 1-minute prediction interval 89
xiv  LIST OF FIGURES

5.9 Short term forecast residual of biphase adaptive differential evolution

algorithm based prediction model on 1-minute prediction interval 89
5.10 Comparing short term forecast errors of SaDE and BaDE based pre-
dictive frameworks 90
5.11 Long term forecast assessment of self-adaptive differential evolution
algorithm based prediction model on 60-minute prediction interval 91
5.12 Long term forecast residuals of self-adaptive differential evolution
algorithm based prediction model on 60-minute prediction interval 91
5.13 Long term forecast assessment of biphase adaptive differential evolu-
tion algorithm based prediction model on 60-minute prediction interval 92
5.14 Long term forecast residuals of biphase adaptive differential evolution
algorithm based prediction model on 60-minute prediction interval 92
5.15 Comparing long term forecast errors of SaDE and BaDE based pre-
dictive frameworks 93
5.16 Comparing forecast accuracy of Maximum, Average, BPNN, SaDE,
BaDE based predictive frameworks on NASA Trace 93
5.17 Comparing forecast accuracy of Maximum, Average, BPNN, SaDE,
BaDE based predictive frameworks on Saskatchewan Trace 94

6.1 Non-directed predictive framework 98

6.2 Self-directed predictive framework 99
6.3 Self-directed learning process 100
6.4 Position update procedures in blackhole algorithm (standard vs. mod-
ified) 101
6.5 Web server workload prediction results of self-directed learning pre-
dictive framework on 5-minute prediction interval 105
6.6 Web server workload prediction residuals auto-correlation of self-
directed learning predictive framework on 5-minute prediction interval 106
6.7 Mean squared error of non-directed and self-directed predictive frame-
works on short term forecasts of web server workloads 107
6.8 Training time (min) of non-directed and self-directed predictive frame-
works on short term forecasts of web server workloads 108
6.9 Cloud server workload prediction results of self-directed learning pre-
dictive framework on 5-minute prediction interval 109
6.10 Cloud server workload prediction residuals auto-correlation of self-
directed learning predictive framework on 5-minute prediction interval 110
6.11 Web server workload prediction results of self-directed learning pre-
dictive framework on 60-minute prediction interval 111
6.12 Web server workload prediction residuals auto-correlation of self-
directed learning predictive framework on 60-minute prediction interval 112
 LIST OF FIGURES  xv

6.13 Mean squared error of non-directed and self-directed predictive frame-

works on long term forecasts of web server workloads 113
6.14 Training time (min) of non-directed and self-directed predictive frame-
works on long term forecasts of web server workloads 114
6.15 Cloud server workload prediction results of self-directed learning pre-
dictive framework on 60-minute prediction interval 115
6.16 Cloud server workload prediction residuals auto-correlation of self-
directed learning predictive framework on 60-minute prediction interval 116

7.1 Decomposition based predictive framework 122

7.2 Decomposition of CPU requests trace 123
7.3 A conceptual view of ensemble stacking 125
7.4 An ensemble of ELMs in workload prediction 127
7.5 Input node selection 128
7.6 Network architecture analysis for short term forecast of decomposition
predictive framework 131
7.7 CPU and Memory data-trace auto-correlation for 5-minute prediction
interval 132
7.8 Short term forecast accuracy of ensemble prediction framework on
CPU Trace 133
7.9 Short term forecast accuracy of ensemble prediction framework on
Memory Trace 134
7.10 Network architecture analysis for long term forecast of decomposition
predictive framework 135
7.11 Long term forecast accuracy of ensemble prediction framework on
CPU Trace 136
7.12 Long term forecast accuracy of ensemble prediction framework on
Memory Trace 136

8.1 An illustration of virtual machine placement scenarios 142

8.2 Resource-efficient load balancing framework design 143
8.3 Chromosome encoding for VM placement 144
8.4 Single point crossover operator for VM placement 145
8.5 Swapping based mutation operator for VM placement 145
8.6 Three cases of VM allocation 149
8.7 Secure and energy-aware load balancing framework design 150
8.8 Power consumption (W) for homogeneous VM requests 152
8.9 Resource utilization for homogeneous VM requests 152
8.10 Presence of conflicting servers (%) for homogeneous VM requests 153
xvi  LIST OF FIGURES

8.11 Power consumption (W) for heterogeneous VM requests 153

8.12 Resource utilization for heterogeneous VM requests 154
8.13 Presence of conflicting servers (%) for heterogeneous VM requests 154
List of Tables

2.1 Friedman test statistics for time series forecasting models 23

2.2 Friedman test ranks for time series forecasting models 23
2.3 Finner test post-hoc analysis of time series forecasting models 23

3.1 Wilcoxon test statistics for error preventive and non-error preventive
time series forecasting model 57
3.2 Finner test post-hoc analysis of error preventive and non-error pre-
ventive time series forecasting models 57

4.1 Friedman test statistics of metaheuristic algorithms based prediction

models 71
4.2 Finner test post-hoc analysis statistics of metaheuristic algorithms
based prediction models (ℵ = 0.05) 73
4.3 Finner test post-hoc analysis results on the null hypothesis of meta-
heuristic algorithms based prediction model 74

5.1 Number of iterations and time elapsed in the training of differential

evolution based predictive models for short term forecasts 90
5.2 Number of iterations and time elapsed in the training of differential
evolution based predictive models for long term forecasts 90
5.3 Iterations elapsed for the training of predictive models using back-
propagation, self-adaptive differential evolution, and biphase adaptive
differential evolution 94
5.4 Training time (sec) elapsed in the training of predictive models us-
ing backpropagation, self-adaptive differential evolution, and biphase
adaptive differential evolution 95

6.1 Mean squared error of non-directed and self-directed predictive frame-

works on short term forecasts of cloud server workloads 110
6.2 Training time (min) of non-directed and self-directed predictive frame-
works on short term forecasts of cloud server workloads 111
6.3 Mean squared error of non directed and self-directed predictive frame-
works on long term forecasts of cloud server workloads 117

xvii
xviii  LIST OF TABLES

6.4 Training time (min) of non-directed and self-directed predictive frame-

works on long term forecasts of cloud server workloads 117
6.5 Mean squared error comparison of non-directed and self-directed learn-
ing based models’ NASA Trace forecasts with state-of-art models 117
6.6 Mean squared error comparison of non-directed and self-directed learn-
ing based models’ Calgary Trace forecasts with state-of-art models 117
6.7 Mean squared error comparison of non-directed and self-directed learn-
ing based models’ Saskatchewan Trace forecasts with state-of-art
models 118
6.8 Mean squared error comparison of non-directed and self-directed learn-
ing based models’ CPU Trace forecasts with state-of-art models 118
6.9 Mean squared error comparison of non-directed and self-directed learn-
ing based models’ Memory Trace forecasts with state-of-art models 118
6.10 Mean squared error comparison of self-directed learning-based model’s
PlanetLab Trace forecasts with deep learning model 118
6.11 Friedman test ranks of non-directed, self-directed, LSTM, SaDE, and
backpropagation based predictive models 119
6.12 Friedman test statistics of non-directed and self-directed learning
predictive frameworks 119
6.13 Wilcoxon signed test ranks for self-directed learning predictive frame-
work 119

7.1 ARIMA analysis orders for cloud resource demand traces 124
7.2 Network configuration parameter choices for decomposition predictive
framework 124
7.3 List of experiments selected by D-Optimal Design 124
7.4 Mean squared error of short term forecast of ELM based predictive
framework 134
7.5 Mean squared error of long term forecast of ELM based predictive
framework 137
7.6 Forecast accuracy comparison of ELM based predictive models on
CPU trace with state-of-art models 138
7.7 Forecast accuracy comparison of ELM based predictive models on
Memory trace with state-of-art models 138
7.8 Forecast accuracy comparison of ELM based predictive models on
Google cluster trace and PlanetLab Trace with state-of-art models 139

8.1 Virtual machine details for illustration 148

Preface

Cloud computing has become one of the revolutionary technology in the history
of the computing world. It offers subscription-based on-demand services and has
emerged as the backbone of the computing industry. It has enabled us to share
resources among multiple users through virtualization by the means of creating a
virtual instance of a computer system running in an abstracted hardware layer. Unlike
early distributed computing models, it assures limitless computing resources through
its large-scale cloud data centers. It has gained wide popularity over the past few
years, with an ever-increasing infrastructure, number of users, and amount of hosted
data. The large and complex workloads hosted on these data centers introduce several
challenges: resource utilization, power consumption, scalability, operational cost, and
many others. Therefore, a practical resource management scheme is essential to bring
operational efficiency with improved elasticity. The elasticity of a system depends on
several factors such as the accuracy of anticipated workload information, performance
behavior of applications in different scenarios communicating the forecast results, use
of the anticipated information, and many others.
Effective resource management can be achieved through workload prediction,
resource scheduling, and provisioning, virtual machine placement, or a combination
of these approaches. The workload prediction has been widely explored and a number
of methods are presented. However, the existing methods suffer from various issues
including the incapability of capturing the non-linearity of workloads and iterative
training that consumes huge computing resources and time. This book discusses
the machine learning-based approaches to address the above-mentioned issues. The
highlights of the discussed models are continuous learning from error feedback, adaptive
nature, decomposition of workload traces, and ensemble learning. Detailed analysis of
predictive methods on different workload traces is also included and their performance
is compared with state-of-art models. Virtual machine placement is another aspect
that is explored to achieve efficient resource management. In general, virtual machine
placement is a multiobjective problem that involves multiple conflicting objectives to
be optimized simultaneously. The frameworks discussed in this book address the issues
of resource utilization, power consumption, and security while placing the workloads
on servers.
The remainder of the book is organized as follows: Chapter 1 briefs the basic cloud
computing concepts. The discussion on the workload prediction mechanisms begins
in chapter 2. First, the basic time series forecasting models are discussed with their
performance on different workload traces. Chapter 3 discusses the error preventive time
series forecasting models which significantly improve the performance over classical
time series models. Then, a discussion on various nature-inspired algorithms is included

xix
xx  Preface

in chapter 4. It also evaluates the performance of neural network-based forecasting

models trained by these algorithms. Next, the forecasting models trained by adaptive
differential evolution are presented in Chapter 5. The first learning algorithm allows
learning the best suitable mutation strategy and crossover rate. In contrast, the second
algorithm allows learning both crossover and mutation strategies along with mutation
and crossover rates. Chapter 6 discusses the blackhole neural network-based forecasting
scheme and evaluates its performance on several workload traces. It also discusses
the concept of self-directed learning. Also, it discusses the self-directed workload
forecasting model inspired by an error preventive scheme along with a modification
in the blackhole learning algorithm to improve the learning capability of the model.
Chapter 7 introduces the decomposition and ensemble learning-based models. The
decomposition-based model trains one network for each component extracted from the
decomposition of workload trace whereas the second approach creates an ensemble
of extreme learning machines and weights their opinions using a blackhole learning
algorithm. Chapter 8 introduces two multi-objective load balancing frameworks. The
first framework considers the resource utilization and power consumption as objectives
to be optimized whereas the second framework also considers the security aspect
into consideration while assigning the VMs to servers. The framework deals with
side-channel attacks only and minimizes the likelihood of the attack occurring. It also
ensures to reduce the number of victim users if any attack occurs. Finally, Chapter 9
summarizes the work discussed in the book.
Author

Dr. Jitendra Kumar is an assistant professor in machine learning at the National

Institute of Technology Tiruchirappalli, Tamilnadu, India. He obtained his doctorate
in 2019 from the National Institute of Technology Kurukshetra, Haryana, India. He is
also a recipient of the Director’s medal for the first rank in the University examination
at Dayalbagh Educational Institute, Agra, Uttar Pradesh in 2011. He has experience
of three years in academia. He has published several research papers in international
journals and conferences of high repute, including IEEE Transactions on Parallel and
Distributed Systems, Information Sciences, Future Generation Computer Systems,
Neurocomputing, Soft Computing, Cluster Computing, IEEE-FUZZ, etc. He has
also obtained the best paper awards in two international conferences. His research
interests are machine learning, cloud computing, healthcare, parallel algorithms, and
optimization. He is also a review board member of several journals, including IEEE
Transactions on Computers, IEEE Transactions on Parallel and Distributed Systems,
IEEE Access, Journal and Parallel Distributed Computing, and more.
Prof. Ashutosh Kumar Singh is an esteemed researcher and academician in
the domain of Electrical and Computer engineering. Currently, he is working as a
Professor; Department of Computer Applications; National Institute of Technology;
Kurukshetra, India. He has more than 20 years of research, teaching, and admin-
istrative experience in various University systems of the India, UK, Australia, and
Malaysia. Dr. Singh obtained his Ph.D. degree in Electronics Engineering from Indian
Institute of Technology-BHU, India; Post Doc from Department of Computer Science,
University of Bristol, UK and Charted Engineer from UK. He is the recipient of
the Japan Society for the Promotion of Science (JSPS) fellowship for a visit to the
University of Tokyo and other universities of Japan. His research area includes Verifi-
cation, Synthesis, Design, and Testing of Digital Circuits, Predictive Data Analytics,
Data Security in Cloud, Web Technology. He has more than 250 publications till
now which includes peer-reviewed journals, books, conferences, book chapters, and
news magazines in these areas. He has co-authored eight books including ‘‘Web Spam
Detection Application using Neural Network’’, ‘‘Digital Systems Fundamentals’’ and
‘‘Computer System Organization & Architecture’’. Prof. Singh has worked as principal
investigator/investigator for six sponsored research projects and was a key member on
a project from EPSRC (United Kingdom) entitled ’’Logic Verification and Synthesis
in New Framework’’.
Dr. Singh has visited several countries including Australia, United Kingdom, South
Korea, China, Thailand, Indonesia, Japan, and the USA for collaborative research
work, invited talks, and present his research work. He had been entitled to 15 awards
such as Merit Awards-2003 (Institute of Engineers), Best Poster Presenter-99 in 86th

xxi
xxii  Author

Indian Science Congress held in Chennai, INDIA, Best Paper Presenter of NSC’99
INDIA and Bintulu Development Authority Best Postgraduate Research Paper Award
for 2010, 2011, 2012.
He has served as an Editorial Board Member of International Journal of Networks
and Mobile Technologies, International Journal of Digital Content Technology and
its Applications. Also, he has shared his experience as a Guest Editor for Pertanika
Journal of Science and Technology, Chairman of CUTSE International Conference
2011, Conference Chair of series of International Conference on Smart Computing and
Communication (ICSCC), and as an editorial board member of UNITAR e-journal.
He is involved in reviewing processes in different journals and conferences of repute
including IEEE transaction of computer, IET, IEEE conference on ITC, ADCOM,
etc.
Prof. Anand Mohan has nearly 44 years of experience in teaching and research
and the administration and management of higher educational institutions. He is
currently an institute professor in the Department of Electronics Engineering, Indian
Institute of Technology (BHU), Varanasi, India. Besides his present academic assign-
ment, Prof. Mohan is a Member of the Executive Council of Banaras Hindu University
and Vice-Chairman of the Board of Governors of Indian Institute of Technology
(BHU), Varanasi, India. Prof. Mohan served as Director (June 2011-June 2016) of
the National Institute of Technology (NIT), Kurukshetra, Haryana, India, and was
also Mentor Director of the National Institute of Technology, Srinagar, Uttarakhand,
India. For his outstanding contributions in the field of Electronics Engineering, Prof.
Mohan was conferred the ’’Lifetime Achievement Award’’ (2016) by Kamla Nehru
Institute of Technology, Sultanpur, India.
Prof. Rajkumar Buyya is a Redmond Barry Distinguished Professor and Direc-
tor of the Cloud Computing and Distributed Systems (CLOUDS) Laboratory at the
University of Melbourne, Australia. He is also serving as the founding CEO of Manjra-
soft Pty Ltd., a spin-off company of the University, commercializing its innovations in
Cloud Computing. He served as a Future Fellow of the Australian Research Council
during 2012-2016. He serving/served as an Honorary/Visiting Professor for several
elite Universities including Imperial College London (UK), University of Birmingham
(UK), University of Hyderabad (India), and Tsinghua University (China). He received
B.E and M.E in Computer Science and Engineering from Mysore and Bangalore
Universities in 1992 and 1995 respectively; and a Doctor of Philosophy (Ph.D.) in
Computer Science and Software Engineering from Monash University, Melbourne,
Australia in 2002. He was awarded Dharma Ratnakara Memorial Trust Gold Medal
in 1992 for his academic excellence at the University of Mysore, India. He received
Richard Merwin Award from the IEEE Computer Society (USA) for excellence in
academic achievement and professional efforts in 1999. He received Leadership and
Service Excellence Awards from the IEEE/ACM International Conference on High-
Performance Computing in 2000 and 2003. He received the ‘‘Research Excellence
Awards’’ from the University of Melbourne for productive and quality research in
computer science and software engineering in 2005 and 2008. With over 112,400
citations, a g-index of 322, and an h-index of 145, he is the highest cited computer
scientist in Australia and one of the world’s Top 30 cited authors in computer science
 Author  xxiii

and software engineering. He received the Chris Wallace Award for Outstanding
Research Contribution 2008 from the Computing Research and Education Association
of Australasia, CORE, which is an association of university departments of computer
science in Australia and New Zealand. Dr. Buyya received the ‘‘2009 IEEE TCSC
Medal for Excellence in Scalable Computing’’ for pioneering the economic paradigm for
utility-oriented distributed computing platforms such as Grids and Clouds. He served
as the founding Editor-in-Chief (EiC) of IEEE Transactions on Cloud Computing
(TCC). Dr. Buyya is recognized as a ‘‘Web of Science Highly Cited Researcher’’ for
five consecutive years since 2016, a Fellow of IEEE and Scopus Researcher of the
Year 2017 with Excellence in Innovative Research Award by Elsevier, and ‘‘Lifetime
Achievement Award’’ from two Indian universities for his outstanding contributions
to Cloud computing and distributed systems. He has been recently recognized as the
‘‘Best of the Worl’’, in the Computing Systems field, by The Australian 2019 Research
Review.
Dr. Buyya has authored/co-authored over 850 publications. Since 2007, he received
twelve ‘‘Best Paper Awards’’ from international conferences/journals including a
‘‘2009 Outstanding Journal Paper Award’’ from the IEEE Communications Society,
USA. He has co-authored five text books: Microprocessor x86 Programming (BPB
Press, New Delhi, India, 1995), Mastering C++ (McGraw Hill Press, India, 1st
edition in 1997 and 2nd edition in 2013), Object Oriented Programming with Java:
Essentials and Applications (McGraw Hill, India, 2009), Mastering Cloud Computing
(Morgan Kaufmann, USA; McGraw Hill, India, 2013; China Machine Press, 2015),
and Cloud Data Centers and Cost Modeling (Morgan Kaufmann, USA, 2015). The
books on emerging topics that he edited include High Performance Cluster Computing
(Prentice Hall, USA, 1999), High Performance Mass Storage and Parallel I/O (IEEE
and Wiley Press, USA, 2001), Content Delivery Networks (Springer, Germany, 2008),
Market Oriented Grid and Utility Computing (Wiley Press, USA, 2009), and Cloud
Computing: Principles and Paradigms (Wiley, USA, 2011). He also edited proceedings
of over 25 international conferences published by prestigious organizations, namely
the IEEE Computer Society Press (USA) and Springer Verlag (Germany). He served
as Associate Editor of Elsevier’s Future Generation Computer Systems Journal (2004-
2009) and currently serving on editorial boards of many journals including Software:
Practice and Experience (Wiley Press). Dr. Buyya served as a speaker in the IEEE
Computer Society Chapter Tutorials Program (from 1999-2001), Founding Co-Chair
of the IEEE Task Force on Cluster Computing (TFCC) from 1999-2004, and member
of the Executive Committee of the IEEE Technical Committee on Parallel Processing
(TCPP) from 2003-2011. He served as the first elected Chair of the IEEE Technical
Committee on Scalable Computing (TCSC) during 2005-2007 and played a prominent
role in the creation and execution of several innovative community programs that
propelled TCSC into one of the most successful TCs within the IEEE Computer
Society. In recognition of these dedicated services to the computing community
over a decade, the President of the IEEE Computer Society presented Dr. Buyya a
Distinguished Service Award in 2008.
Dr. Buyya has contributed to the creation of high-performance computing and
communication system software for PARAM supercomputers developed by the Centre
xxiv  Author

for Development of Advanced Computing (C-DAC), India. He has pioneered Economic

Paradigm for Service-Oriented Distributed Computing and demonstrated its utility
through his contribution to conceptualization, design, and development of Grid and
Cloud Computing technologies such as Aneka, GridSim, Libra, Nimrod-G, Gridbus,
and Cloudbus that power the emerging eScience and eBusiness applications. He has
been awarded, over $8 million, competitive research grants from various national and
international organizations including the Australian Research Council (ARC), Sun
Microsystems, StorageTek, IBM, and Microsoft, CA Australia, Australian Department
of Innovation, Industry, Science, and Research (DIISR), and European Council. Dr.
Buyya has been remarkably productive in a research sense and has converted much
of that knowledge into linkages with industry partners (such as IBM, Sun, and Mi-
crosoft), into software tools useful to other researchers in a variety of scientific fields,
and into community endeavors. Software technologies for Grid and Cloud computing
developed under Dr. Buyya’s leadership have gained rapid acceptance and are in use
at several academic institutions and commercial enterprises in 50+ countries around
the world. In recognition of this, he received Vice Chancellor’s inaugural ‘‘Knowledge
Transfer Excellence (Commendation) Award’’ from the University of Melbourne in
Nov 2007. Manjrasoft’s Aneka technology for Cloud Computing developed under
Dr.Buyya’s leadership has received the ‘‘2010 Asia Pacific Frost & Sullivan New Prod-
uct Innovation Award’’. Recently, Dr. Buyya received the ‘‘Bharath Nirman Award’’
and the ‘‘Mahatma Gandhi Award’’ along with Gold Medals for his outstanding and
extraordinary achievements in Information Technology field and services rendered to
promote greater friendship and India-International cooperation.
Abbreviations

SLA Service Level Aggrement

QoS Quality of Service
DE Differential Evolution
PSO Particle Swarm Optimization
EA Evolutionary Algorithm
FSA Firefly Search Algorithm
HS Harmony Search
TLBO Teaching Learning Based Optimization
GSA Gravitational Search Algorithm
BhOA Blackhole Algorithm
BhNN Bloackhole Network
MSE Mean Squared Error
RMSE Root Mean Squared Error
MAE Mean Absolute Error
RMAE Relative Mean Absolute Error
CoC Correlation Coefficient
SEI Sum of Elasticity Index
EP Error Preventive
EPS Error Prevention Score
NEP Non Error Preventive
PER Predictions in Error Range
MoP Magnitude of Prediction
SaDE Self Adaptive Differential Evolution
BaDE Biphase Adaptive Differential Evolution
SDL Self Directed Learning
ELM Extreme Learning Machine
RELB Resource Efficient Load Balancing
SCA Side Channel Attack
SEALB Secure and Energy Aware Load Balancing
PWS Prediction Window Size
WPBPNN BPNN based Workload Prediction Model
WPBhNN BhNN based Workload Prediction Model
WPSDL
BhNN BhNN and SDL based Workload Prediction Model
WPSaDE SaDE based Workload Prediction Model
WPBaDE BaDE based Workload Prediction Model
ELMNN ELM based Neural Network

xxv
xxvi  Abbreviations

eELMNN Ensemble of ELM based Neural Networks

WPELMNN ELMNN based Workload Prediction Model
WPeELMNN eELMNN based Workload Prediction Model
N DS Non Dominated Sorting
 CHAPTER 1

Introduction

loud computing paradigm enables the delivery of computing resources and

C applications to users across the globe as subscription-oriented services. Virtu-
alization is the technique behind the scene that helps in resource sharing among
multiple users in this cloud computing environment.

1.1 CLOUD COMPUTING

Cloud computing is a form of distributed computing environment where multiple
virtual instances of a computer system run in abstracted hardware level and every user
experiences like owning the entire system. The cloud infrastructure may be private
(serves to a single organization), public (shared among multiple organizations), and
hybrid (combination of both). A cloud system provides the on-demand services at
three different levels, referred to as Infrastructure as a Service (IaaS), Platform as a
Service (PaaS), and Software as a Service (SaaS), as shown in Fig. 1.1. In IaaS, the
infrastructure components such as servers, networking and storage, and operating ser-
vices hosted by service providers are delivered to the consumers through virtualization.
These components are provided with various services including monitoring, security,
log access, backup and recovery, and load balancing. While in PaaS, users get required
and associated infrastructure to develop, run, and manage their applications. The
service provider is responsible for providing the servers, operating system, storage,
database, and middleware such as Java and .NET runtime. In the case of SaaS
that is a software distribution model, the software or applications are hosted in the
data centers, and users access these applications over the Internet. The applications
delivered through SaaS eliminate the requirement of hardware, installation, support,
and maintenance as they do not need any installation on local computers and can be
accessed through web browsers.
In the last decade, cloud systems have received wide popularity due to ever-
growing services, infrastructure, clients, and the ability to host big data [17]. A
survey conducted in 2017 reported that organizations would shift their 90% enterprise
workload on a cloud by 2021 [29]. The cloud infrastructure is growing very fast, and
the cloud industry is expected to grow with 14.6% compound annual growth rate
to reach the $300 billion mark by 2022 [62, 63]. Modern cloud systems are equipped
with characteristics such as on-demand service, reliability, scalability, elasticity,

DOI: 10.1201/9781003110101-1 1
2  Machine Learning for Cloud Management

SaaS

User Ap-
plications
PaaS

Database Application
UI Services
Grid Grid
IaaS

Virtual Virtual Virtual Virtual Virtual

Network Machine Storage Machine Server

Amazon Elastic Compute

Google Apps, Drop- Amazon Web Services,
Cloud, Microsoft Azure,
box, Salesforce, etc. Google Compute Engine, etc.
Google Apps Engine, etc.

End Users Software Developers Infrastructure & Network Architects

FIGURE 1.1 Service model view of cloud computing

disaster recovery, accessibility, measured services, and many others [25, 73, 75, 98].
However, various challenges and limitations are still open including inefficient resource
management, security and privacy, heterogeneity, elasticity, usability, response time,
and many more [18, 19, 21, 52--54, 90, 109, 124, 125].

Resource
Clients Information

Web
Portal
Resource Management
Resource Pool

Workload
Workload Resource
Analysis Provision
Cloud Scheduler

FIGURE 1.2 Cloud resource management view

1.2 CLOUD MANAGEMENT

Resource management is one of the core functions of cloud systems and must be
improved for better system performance [66, 102, 119]. The inefficiency in resource
management directly affects the system performance and operational cost. The poor
resource utilization degrades the overall system performance and may increase the
service cost as well. A simple resource management block diagram in a cloud system is
depicted in Fig. 1.2. It can be seen that clients are connected to a cloud server through
a web portal. Users send their workloads to the cloud server for the execution. In turn,
 Introduction  3

a modern cloud system tries to assign the workloads to one of the server machines
based on different criteria including resource utilization, system performance, user
priorities, operational cost, quality of service, etc. Typically, the complete process of
workload placement over a time to improve different variables of a system is referred
to as cloud resource management. As depicted in Fig. 1.2, the major tasks of a cloud
resource management application are workload analysis and forecasting, resource
provisioning, and scheduling the workloads on hardware. The workload analysis
module is responsible for analyzing the upcoming workload and for forecasting the
expected workload in the near future. This information is used by the resource
provisioning module to allocate the physical resources. The resource scheduler places
the workloads on the servers based on the input from the resource provisioning module
and current resource usage information. Typically, resource management is achieved
through prediction, scaling, provisioning, and load balancing, as shown in Fig. 1.3.
However, this book concentrates on workload forecasting using different approaches
of regression analysis and artificial neural networks, and load balancing.

Cloud Resource
Management

Workload Application Scaling and

Load Balancing
Prediction Provisioning

Resource Demand Resource Utilization Scheduling VM Placement

FIGURE 1.3 Cloud resource management approaches

1.2.1 Workload Forecasting

The workload prediction is a mechanism that estimates the future workload on the
servers and can be classified as either a homeostatic or history-based method [89].
The first class of methods detect a trend in previous actual values and add or subtract
it to the current value to forecast the next value. It could be a static or dynamic
value that is to be added/subtracted. On the other hand, the second class of models
analyzes the workload history and extracts a pattern to forecast the next value. A
homeostatic method attempts to follow the mean of the previous values, while the
history-based approach uses the behavior of previous workload information to forecast
the next instance [70,114,116−118].
4  Machine Learning for Cloud Management

Learning Window Forecast

x1 x2 ... xt x̂t+1

Forecaster

FIGURE 1.4 Schematic representation of workload forecasting

x̂t+1 = f (xt , xt−1 , . . . , x1 ) (1.1)

Let f be a function of X which determines the value of x̂t+1 i.e. the estimated
workload at time t + 1. In order to forecast the upcoming workload, the f analyses the
historical workloads of the length of the learning window (eq. 1.1). For instance, the
function analyses the previous 10 instances from history if the length of the learning
window is 10. Since the forecast function of a real data-trace is generally non-linear
and complex, it becomes a challenging task to find the set of optimal parameters.
And machine learning becomes the natural choice to optimize the model parameters
to forecast the dynamic and non-linear workloads. A typical workload forecasting
model is depicted in Fig. 1.4. The learning window defines the number of recent past
workload instances to be analyzed for anticipation of next the value.
The prediction models have been explored and developed for various applica-
tions [13, 37, 46, 96, 97, 99]. The workload predictive resource management approaches
are tailored with estimations of demand and utilization of resources. The workload
of cloud services is dynamic and varies over time [30, 100]. Therefore a robust pre-
diction model is required to produce reasonably accurate forecasts. On the other
hand, the resource utilization prediction helps in accessing the free resources and
also in accessing the impact of allocating the free resources to individual work-
loads [35, 76, 77, 79, 80, 115, 121, 123, 134].

1.2.2 Load Balancing

The task of a load balancing process is to distribute the workloads uniformly among
servers. The load balancers are responsible for identifying the best suitable servers
or computing resources that meet the application requirements. It ensures that the
high volume of network traffic is not diverted to a single server. The schematic
representation of load balancing in a distributed computing environment is shown in
Fig. 1.5. The load balancer receives the traffic of users’ requests through the Internet
and distributes it among accessible and eligible servers. For instance, when seven users
send their workloads, the load balancer balances the load distribution by assigning
the load to three of the four servers.
 Introduction  5

Cloud Datacenter

Se Se
rve rve
r r

Se Se
rve rv
r er

WWW Load
Balancer
Cloud Users

FIGURE 1.5 Load balancing

The effective load balancing is another approach that helps in achieving better
usage of resources and their management. The efficiency of load balancing approaches
has been an issue for cloud systems since its development [84, 129]. The efficiency in
load balancing can be achieved using different approaches such as optimal scheduling
and placement of workloads or virtual machines (VMs). The optimal mapping of
VMs is a complex and challenging task as it involves multiple objectives to optimize
at the same time and belongs to NP-Complete class of problems [16, 88]. Generally,
the existing VM placement algorithms consider the different dimensions of resource
utilization and power consumption in the data centers [4, 138]. We will focus on the
load balancing approaches, also dealing with the security while balancing the load on
cloud servers as it is one of the most important issues in the cloud architectures, and
various approaches have been discovered including [43, 85].

1.3 MACHINE LEARNING

Machine learning allows a computer to master a specific task without being explicitly
programmed. A computer can extract the underlying rules to perform the given task
from a bunch of data points. In this book, we will discuss the cloud management
models which are developed using the following techniques:

1.3.1 Artificial Neural Network

An artificial neural network (ANN) is composed of huge interconnected nodes called
neurons, as shown in Fig. 1.6, that processes any information in a similar way as
of human brain. Typically, a neural network is trained to solve a specific complex
problem such as recognition, classification, forecasting, and others. Similar to the
brain, a neural network adjusts the connections and their weights during the learning
phase. The neural networks are capable of analyzing the complex and large amount
of data and extracting the patterns from it.
6  Machine Learning for Cloud Management

Output
Layer

Input Hidden Hidden

Layer Layer-3 Layer-4

Hidden Hidden
Layer-1 Layer-2

FIGURE 1.6 Artificial neural network

The key difference between a traditional computing approach and a neural network
is that the traditional approach follows a set of rules that must be known to the
computer in advance, while a neural network can learn from the data itself to draw
insightful inferences using some specific rules. Let κ1 = [x1 , x2 , . . . , xt ] be an input
vector, and the network, as shown in Fig. 1.6, is applied to estimate the value of xt+1 .
Assuming that ωi,j k
represents the weight of a synaptic connection between the ith
node of the k th layer and the j th node of the next layer, and ζk denotes the activation
function applied on k th layer nodes. The output of the j th node of layer k + 1 can be
computed as zj = ti=1 ζk+1 (xi · ωi,jk
) that acts as the input to next layer nodes.
P

1.3.2 Metaheuristic Optimization Algorithms

Optimization has become an integral part of solving real-world problems which are
multi-modal and highly non-linear in nature. These problems can be represented
as a constrained optimization problem with one or more decision variables. Some
of the real-world optimization problems are routing, engineering designs, resource
assignment, and most of these problems are NP-hard [16, 86]. Consider that Fig. 1.7
shows an arbitrary function farb of decision variable x that needs to be minimized.
The solution space has an infinite number of solutions along with multiple local
optima such as Sl1 , Sl2 , Sl3 , and many others. If the solution space is multidimensional,
complex, and large enough, it becomes a challenging task to find the global optimal
(Sg∗ ) in a reasonable time. The metaheuristic optimization algorithm helps in solving
such problems. The term metaheuristic was first used by Fred Glover [47] for an
approach that has the capability of guiding and modifying the other heuristics to
produce the solutions beyond their ability [48]. A metaheuristic algorithm does not
guarantee to produce an optimal solution, but it generates an approximated solution
in a reasonable amount of time.
 Introduction  7

farb

Sl1 Sl2 Sg∗ Sl4

Decision Variable (x)

FIGURE 1.7 An arbitrary optimization function with multiple local optima

These algorithms can be classified into two major categories i.e. trajectory-based
and population-based approaches. A trajectory-based algorithm such as Simulated
Annealing works around a single solution to find an optimal solution for the problem
under consideration. On the other hand, a population-based algorithm uses a set
of solutions to search for an optimal solution. A detailed study on metaheuristic
optimization can be seen in [14].

1.3.3 Time Series Analysis

A sequence of data points obtained at regular interval and indexed in time is referred
to as time series data. Time series analysis started a long ago in 1927 [128] and has
a range of applications including finance, signal processing, astronomy, forecasting,
stock market, statistics, defense, politics, etc. In general, the task of a time series
analysis model is to extract the meaningful data characteristics to predict the data
trends. The time series models predict the future event after analyzing the historical
events, and a number of models are introduced. Time series analysis is applicable to
any kind of data including numeric, symbolic, continuous, and real-valued data.

1.4 WORKLOAD TRACES

The analysis reported in this book is carried out on various data traces. The workload
data traces belong to two different categories, namely web server workloads and cloud
server workloads.

HTTP-Web Server Logs: The HTTP traces of web servers of NASA, Calgary, and
Saskatchewan servers are used [1]. In this book, these data traces are referred to as
NASA Trace (D1 ), Calgary Trace (D2 ), Saskatchewan Trace (D3 ), respectively. The
D1 is composed of two months of HTTP web requests obtained from the WWW server
of NASA Kennedy Space Center in Florida. Similarly, the D2 data-trace contains the
HTTP request of one-year duration obtained from the WWW server located at the
University of Calgary, Alberta, Canada. On the other hand, the D3 is a data-trace
that contains the HTTP server requests of seven months obtained from a WWW
8  Machine Learning for Cloud Management

server of a university at Saskatchewan. Every data trace stores the records in ASCII
files, and every line stores one record. Every record is composed of five records i.e.
host, timestamp, request, HTTP reply code, and bytes in the reply.

Google Cluster Trace: It contains the data collected from the cluster cell of Google
for 29 days of duration. The workload trace was released in 2011, and it contains
the data from 10388 servers, 20 million tasks, and more than 0.67 million jobs [112].
A job is a set of one or more tasks, and tasks are further decomposed into one or
more processes. In this book, the CPU and Memory resource demands are used and
referred to as CPU Trace (D4 ) and Memory Trace (D5 ).

PlanetLab Trace: It is a collection of mean CPU utilization data which is collected

from 11,746 virtual machines. These virtual machines are scattered at 500 different
locations across the world. The data was collected for randomly selected 10 days
during March and April of 2011, and data was sampled on five minutes intervals. The
CPU utilization data of 22 randomly selected virtual machines is used in this book
and referred to as PlanetLab Trace (D6 ).

1.5 EXPERIMENTAL SETUP & EVALUATION METRICS

A machine equipped which contains two Intel® Xeon® E5-2630 v4 processors, and
both processors run at the clock speed of 2.20 GHz. The machine is equipped with
128GB of main memory, and it operates on 64-bit windows servers 2012 R2. The
predictive frameworks discussed in this book are evaluated using below mentioned
metrics:

Mean Squared Error: The mean squared error (MSE) measures the forecast accuracy,
and it is one of the popular metrics used in the literature. This method heavily penalizes
the large error terms. Mathematically, it is denoted as given in eq. (1.2), where m
represents the size of data in a given trace. The term MSE and MPE (mean squared
prediction error) are interchangeably used in the book. Moreover, the square root of
MSE (RMSE) may also be used as an error metric.
m
1 X
M SE = (xt − x̂t )2 (1.2)
m t=1

Mean Absolute Error: A small number of very large magnitude errors may influence
the accuracy measured using mean squared error. Whereas mean absolute error equally
weights every error term, and it computes the mean of absolute differences between
predicted and actual workloads as given in eq. (1.3). The forecasts are close to the
actual workload values if the measured score is close to zero.
m
1 X
M AE = |xt − x̂t | (1.3)
m t=1
 Introduction  9

Relative Mean Absolute Error: A scale-free error metric is required to compare the
forecast models on different data sets, and relative mean absolution error (RelMAE)
is one such metric. The score can be calculated using eq. (1.4), which represents the
mean absolute error of the algorithm (M AEA ) normalized by the mean absolute error
of a base or state of the art model (M AEBM )

M AEA
RelM AE = (1.4)
M AEBM

Mean Absolute Scaled Error: Rob J. Hyndman and Anne B. Koehler introduced a
new metric as a substitution of percentage error metrics [61]. The prediction errors
are scaled on the basis of the training mean absolute error of a naı̈ve forecast method.
It computes the measured score using eq. (1.5), where ms denotes the seasonal term.
This metric is a good choice of accuracy measurement when the prediction model is
compared across a number of different scales.
m
1 X
!
|xt − x̂t |
M ASE(x, x̂) = 1 Pm (1.5)
m t=1 m−ms t=ms +1 |xt − xt−1 |

Correlation Coefficient: The correlation coefficient (CoC) statistically evaluates the

statistical relationship of two variables by measuring the degree of movements. The
¯ are the mean values
CoC score can be computed as given in eq. (1.6), where x̄ and x̂
of actual and predicted workloads, respectively.
P ¯
(xt − x̄)(x̂t − x̂)
CoCx x̂ = qP (1.6)
(xt − x̄)2
P ¯ 2
(x̂t − x̂)

Sum of Elasticity Index: Messias et al. proposed to use the sum of elasticity index
(SEI) as a measure of forecast accuracy [92]. This metric supports a forecast model
having the best performance most of the time. As opposed to MAE and RMSE, it is
very less sensitive to the outliers. The SEI score is computed as given in eq. (1.7) and
it always lies between zero and one, where zero and one define the worst and best
accuracy of the model.
m
min(xt , x̂t )
SEI = (1.7)
X

t=1
max(xt , x̂t )

1.6 STATISTICAL TESTS

The statistical techniques are used to analyze the forecasting behavior of different
approaches. The non-parametric tests are used due to the fact that they are not highly
restrictive and can be used over small sample sizes [44]. The significance tests help in
finding the presence of significant differences in two or more forecasting models.
10  Machine Learning for Cloud Management

1.6.1 Wilcoxon Signed-Rank Test

The Wilcoxon signed-rank test is one of the non-parametric tests that compare two
samples to find out whether they represent the same population or not [130]. It
assumes a null hypothesis (H0WC ) that the mean of both samples is the same. If a
significant difference is detected, H0WC gets rejected.
Considering that two algorithms are being evaluated on k problems and i denotes
the performance score difference on the ith problem. The method ranks the absolute
differences and computes the ranks accordingly. The ties can be addressed using one
of the available approaches. In this book, the number of ties is equally divided to
compute the rank of both algorithms. The total rank of the first algorithm where it
outperforms the second is computed using eq. (1.8) whereas eq. (1.9) computes the
sum of ranks for the problems where the second algorithm gives better results than
the first algorithm.
1X
R+
WC = rank(i ) + rank(i ) (1.8)
X

i >0
2 i =0
1X
WC =
R− rank(i ) + rank(i ) (1.9)
X

i <0
2 i =0

1.6.2 Friedman Test

It is a non-parametric test developed by M. Friedman [40, 41] that provides an
alternative to one-way ANOVA with repeated measures. It conducts multiple tests
that target to detect the presence of differences between the performance behavior of
two or more models [34].
Let H0FR be the null hypothesis of Friedman test that states the equality in the
mean of every prediction model’s result. The alternate hypothesis (H1FR ) of the test
is the negation of H0FR . First, it converts the original results of each algorithm into
j
ranks. Let RFR
i
be the Friedman rank of j th algorithm on the ith problem, the final
j
rank of j th algorithm can be observed by calculating the average of RFR i
as shown in
eq. (1.10), where j = {1, 2, . . . , k} and i = {1, 2, . . . , |D|} denote the algorithms and
datasets respectively, |D| is the number of datasets. The minimum value of ranks
represents the best algorithm.
|D|
j j
RFR = (1.10)
X
i
RFR
i=1

1.6.3 Finner Test

The Friedman test conducts the multiple comparison test and detects the significant
difference over the whole population test. However, it is unable to conduct comparisons
to detect the difference between some of the algorithms. Post-hoc analysis tests deliver
the purpose and allow to detect the presence of difference in the performance of
two algorithms on the basis of a control method [38]. The test adjusts the value of
 Introduction  11

significance level (ℵ) in a step-down manner [34]. Considering that the generated
p-values are sorted in an increasing fashion in such a way that pi ≤ pi+1 ; ∀i =
{1, 2, . . . , k − 2}. Let HiFN be the corresponding hypothesis for tests. The Finner test
rejects the hypothesis from H1FN to Hi−1 FN
provided i is the smallest integer number
k−1
that satisfy pi > 1 − (1 − ℵ) i property [34].
References
Traces available in the internet traffic archive . ftp://ita.ee.lbl.gov/html/. [Online: accessed on 01-05-2019].
Oludare Isaac Abiodun , Aman Jantan , Abiodun Esther Omolara , Kemi Victoria Dada , Abubakar Malah Umar
, Okafor Uchenwa Linus , Humaira Arshad , Abdullahi Aminu Kazaure , Usman Gana , and Muhammad Ubale
Kiru . Comprehensive review of artificial neural network applications to pattern recognition. IEEE Access,
7:158820–158846, 2019.
Joseph Ackora-Prah , Samuel Asante Gyamerah, and Perpetual Saah Andam. A heuristic crossover for
portfolio selection. 8(65):3215–3227, 2014.
Fares Alharbi , Yu-Chu Tian , Maolin Tang , Wei-Zhe Zhang , Chen Peng , and Minrui Fei . An ant colony
system for energy-efficient dynamic virtual machine placement in data centers. Expert Systems with
Applications, 120:228–238, 2019.
Osama Moh'd Alia and Rajeswari Mandava . The variants of the harmony search algorithm: an overview.
Artificial Intelligence Review, 36(1):49–68, Jun 2011.
Maryam Amiri and Leyli Mohammad-Khanli . Survey on prediction models of applications for resources
provisioning in cloud. Journal of Network and Computer Applications, 82:93–113, 2017.
E. Assareh , M.A. Behrang , M.R. Assari , and A. Ghanbarzadeh . Application of PSO (particle swarm
optimization) and GA (genetic algorithm) techniques on demand estimation of oil in iran. Energy,
35(12):5223–5229, 2010.
Rasoul Azizipanah-Abarghooee , Taher Niknam , Farhad Bavafa , and Mohsen Zare . Short-term scheduling of
thermal power systems using hybrid gradient based modified teaching-learning optimizer with black hole
algorithm. Electric Power Systems Research, 108:16–34, 2014.
Indu Bala and Anupam Yadav . Gravitational search algorithm: A state-of-the-art review. In Neha Yadav ,
Anupam Yadav , Jagdish Chand Bansal , Kusum Deep , and Joong Hoon Kim , editors, Harmony Search and
Nature Inspired Optimization Algorithms, pages 27–37, Singapore, 2019. Springer Singapore.
F. Baldan , S. Ramirez-Gallego , C. Bergmeir , F. Herrera , and J. M. Benitez . A forecasting methodology for
workload forecasting in cloud systems. IEEE Transactions on Cloud Computing, 6(04):929–941, oct 2018.
Anton Beloglazov and Rajkumar Buyya . Optimal online deterministic algorithms and adaptive heuristics for
energy and performance efficient dynamic consolidation of virtual machines in cloud data centers. Concurrency
Computing: Pract. Exper., 24(13):1397–1420, Sep 2012.
P. D. Bharathi , P. Prakash , and M. V. K. Kiran . Virtual machine placement strategies in cloud computing. In
2017 Innovations in Power and Advanced Computing Technologies (i-PACT), pages 1–7, April 2017.
Mamun Bin Ibne REAZ and Mohd Marufuzzaman . Pattern matching and reinforcement learning to predict the
user next action of smart home device usage. Acta Technica Corviniesis - Bulletin of Engineering, 6(3):37–40,
2013.
Ilhem Boussaïd , Julien Lepagnot , and Patrick Siarry . A survey on optimization metaheuristics. Information
Sciences, 237:82–117, 2013.
George E. P. Box , Gwilym M. Jenkins , and Gregory C. Reinsel . Time Series Analysis. John Wiley & Sons,
Inc., 2008.
Wayne F. Boyer and Gurdeep S. Hura . Non-evolutionary algorithm for scheduling dependent tasks in
distributed heterogeneous computing environments. Journal of Parallel and Distributed Computing,
65(9):1035–1046, Sep 2005.
Rajkumar Buyya , James Broberg , and Andrzej M. Goscinski . Cloud Computing Principles and Paradigms.
Wiley Publishing, 2011.
Rajkumar Buyya , Satish Narayana Srirama , Giuliano Casale , Rodrigo Calheiros , and et al. A manifesto for
future generation cloud computing: Research directions for the next decade. ACM Comput. Surv.,
51(5):105:1–105:38, Nov 2018.
Rajkumar Buyya , Chee Shin Yeo , Srikumar Venugopal , James Broberg , and Ivona Brandic . Cloud
computing and emerging IT platforms: Vision, hype, and reality for delivering computing as the 5th utility. Future
Generation Computer Systems, 25:599–616, 6 2009.
H. R. Cai , C. Y. Chung , and K. P. Wong . Application of differential evolution algorithm for transient stability
constrained optimal power flow. IEEE Transactions on Power Systems, 23(2):719–728, May 2008.
K. Chandrasekaran . Essentials of Cloud Computing. Chapman and Hall/CRC, 1 edition, 2014.
Mingzhe Chen , Ursula Challita , Walid Saad , Changchuan Yin , and Mérouane Debbah . Artificial neural
networks-based machine learning for wireless networks: A tutorial. IEEE Communications Surveys Tutorials,
21(4):3039–3071, 2019.
Z. Chen , A. Sinha , and P. Schaumont . Using virtual secure circuit to protect embedded software from side-
channel attacks. IEEE Transactions on Computers, 62(1):124–136, Jan 2013.
Sakshi Chhabra and Ashutosh Kumar Singh . Dynamic data leakage detection model based approach for
mapreduce computational security in cloud. In Proceedings on 5th International Conference on Eco-Friendly
Computing and Communication Systems, ICECCS 2016, pages 13–19, 4 2017.
Sakshi Chhabra and Ashutosh Kumar Singh . Optimal VM placement model for load balancing in cloud data
centers. In 2019 7th International Conference on Smart Computing and Communications, ICSCC 2019, 6 2019.
Sakshi Chhabra and Ashutosh Kumar Singh . A secure VM allocation scheme to preserve against co-resident
threat. International Journal of Web Engineering and Technology, 15:96–115, 2020.
R. B. Cleveland , W. S. Cleveland , J.E. McRae , and I. Terpenning . STL: A seasonal-trend decomposition
procedure based on loess. Journal of Official Statistics, 6:3–73, 1990.
William Jay Conover and William Jay Conover . Practical nonparametric statistics. Wiley New York, 1980.
Credit Suisse . 2018 Data Center Market Drivers: Enablers Boosting Enterprise Cloud Growth.
https://cloudscene.com/news/2017/12/2018-data-center-predictions/, 2017. [Online; accessed 19-05-2019].
M. E. Crovella and A. Bestavros . Self-similarity in world wide web traffic: evidence and possible causes.
IEEE/ACM Transactions on Networking, 5(6): 835–846, Dec 1997.
A. Cui , Y. Luo , and C. Chang . Static and dynamic obfuscations of scan data against scan-based side-channel
attacks. IEEE Transactions on Information Forensics and Security, 12(2):363–376, Feb 2017.
P.F. de Aguiar , B. Bourguignon , M.S. Khots , D.L. Massart , and R. Phan-Than-Luu . D-optimal designs.
Chemometrics and Intelligent Laboratory Systems, 30(2):199–210, 1995.
K. Deb , A. Pratap , S. Agarwal , and T. Meyarivan . A fast and elitist multiobjective genetic algorithm: NSGA-II.
IEEE Transactions on Evolutionary Computation, 6(2):182–197, April 2002.
Joaquín Derrac , Salvador García , Daniel Molina , and Francisco Herrera . A practical tutorial on the use of
nonparametric statistical tests as a methodology for comparing evolutionary and swarm intelligence algorithms.
Swarm and Evolutionary Computation, 1(1):3–18, 2011.
Saber M. Elsayed , Ruhul A. Sarker , and Daryl L. Essam . An improved self-adaptive differential evolution
algorithm for optimization problems. IEEE Transactions on Industrial Informatics, 9:89–99, 2013.
F. Farahnakian , A. Ashraf , T. Pahikkala , P. Liljeberg , J. Plosila , I. Porres , and H. Tenhunen . Using ant
colony system to consolidate vms for green cloud computing. IEEE Transactions on Services Computing,
8(2):187–198, March 2015.
Araf Farayez , Mamun Bin Ibne Reaz , and Norhana Arsad . SPADE: Activity Prediction in Smart Homes Using
Prefix Tree Based Context Generation. IEEE Access, 7:5492–5501, 2019.
H. Finner . On a monotonicity problem in step-down multiple test procedures. Journal of the American
Statistical Association, 88(423):920–923, 1993.
Iztok Fister , Iztok Fister , Xin-She Yang , and Janez Brest . A comprehensive review of firefly algorithms.
Swarm and Evolutionary Computation, 13:34–46, 2013.
Milton Friedman . The use of ranks to avoid the assumption of normality implicit in the analysis of variance.
Journal of the American Statistical Association, 32(200):675–701, 1937.
Milton Friedman . A comparison of alternative tests of significance for the problem of m rankings. The Annals of
Mathematical Statistics, 11(1):86–92, 1940.
Alexander A. Frolov , Dusan Husek , and Pavel Yu. Polyakov . Recurrent-neural-network-based boolean factor
analysis and its application to word clustering. IEEE Transactions on Neural Networks, 20(7):1073–1086, 2009.
G S Smrithy , Alfredo Cuzzocrea , and Ramadoss Balakrishnan . Detecting Insider Malicious Activities in Cloud
Collaboration Systems. Fundamenta Informaticae, 161(3):299–316, Jul 2018.
Salvador García , Daniel Molina , Manuel Lozano , and Francisco Herrera . A study on the use of non-
parametric tests for analyzing the evolutionary algorithms' behaviour: a case study on the CEC'2005 special
session on real parameter optimization. Journal of Heuristics, 15(6):617, May 2008.
Zong Woo Geem , Joong Hoon Kim , and G.V. Loganathan . A new heuristic optimization algorithm: Harmony
search. SIMULATION, 76(2):60–68, 2001.
Salyean Giri , Abeer Alsadoon , Chandana Withana , Salih Ali , and A. Elchouemic . Prediction of dementia by
increasing subspace size in rank forest. In 2018 IEEE 8th Annual Computing and Communication Workshop
and Conference (CCWC), pages 255–260. IEEE, Jan 2018.
Fred Glover . Future paths for integer programming and links to artificial intelligence. Computers & Operations
Research, 13(5):533–549, Jan 1986.
Fred. Glover and Manuel. Laguna . Tabu search. Kluwer Academic Publishers, 1997.
M. Godfrey and M. Zulkernine . Preventing cache-based side-channel attacks in a cloud environment. IEEE
Transactions on Cloud Computing, 2(4):395–408, Oct 2014.
David Goldberg . Genetic algorithms in search, optimization and machine learning. MA: Addison-Wesley
Professional, ISBN 978-0201157673.
Guang-Bin Huang , Qin-Yu Zhu , and Chee-Kheong Siew . Extreme learning machine: a new learning scheme
of feedforward neural networks. In 2004 IEEE International Joint Conference on Neural Networks (IEEE Cat.
No.04CH37541), volume 2, pages 985–990. IEEE.
Ishu Gupta , Rishabh Gupta , Ashutosh Kumar Singh , and Rajkumar Buyya . MLPAM: A machine learning and
probabilistic analysis based model for preserving security and privacy in cloud environment. IEEE Systems
Journal, 2020.
Ishu Gupta and Ashutosh Kumar Singh . A confidentiality preserving data leaker detection model for secure
sharing of cloud data using integrated techniques. In 2019 7th International Conference on Smart Computing
and Communications, ICSCC 2019. Institute of Electrical and Electronics Engineers Inc., 6 2019.
Ishu Gupta , Niharika Singh , and Ashutosh Kumar Singh . Layer-based privacy and security architecture for
cloud data sharing. Journal of Communications Software and Systems, 15:173–185, 6 2019.
Y. Han , J. Chan , T. Alpcan , and C. Leckie . Using virtual machine allocation policies to defend against co-
resident attacks in cloud computing. IEEE Transactions on Dependable and Secure Computing, 14(1):95–108,
Jan 2017.
Ahmad Hassanat and Esra' Alkafaween . On enhancing genetic algorithms using new crossovers. arXiv
preprint arXiv:1801.02335, 2018.
Abdolreza Hatamlou . Black hole: A new heuristic optimization approach for data clustering. Information
Sciences, 222:175–184, 2013.
John H Holland . Adaptation in natural and artificial systems: An introductory analysis with applications to
biology, control, and artificial intelligence. U Michigan Press, Oxford, England, 1975.
Guang-Bin Huang , Qin-Yu Zhu , and Chee-Kheong Siew . Extreme learning machine: Theory and applications.
Neurocomputing, 70:489–501, 2006.
Rob J Hyndman and Yeasmin Khandakar . Automatic time series forecasting: the forecast package for R.
Journal of Statistical Software, 26(3):1–22, 2008.
Rob J. Hyndman and Anne B. Koehler . Another look at measures of forecast accuracy. International Journal of
Forecasting, 22(4):679–688, 2006.
IDC . Cloud IT Infrastructure Revenues Surpassed Traditional IT Infrastructure Revenues for the First Time in
the Third Quarter of 2018, According to IDC. https://www.idc.com/getdoc.jsp?containerId=prUS44670519,
2019. [Online; accessed 19-05-2019].
Gartner Inc . Gartner Forecasts Worldwide Public Cloud Revenue to Grow 17.3 Percent in 2019.
https://www.gartner.com/en/newsroom/press-releases/2018-09-12-gartner-forecasts-worldwide-public-cloud-
revenue-to-grow-17-percent-in-2019, 2018. [Online; accessed 19-05-2019].
S. M. Islam , S. Das , S. Ghosh , S. Roy , and P. N. Suganthan . An adaptive differential evolution algorithm
with novel mutation and crossover strategies for global numerical optimization. IEEE Transactions on Systems,
Man, and Cybernetics, Part B (Cybernetics), 42(2):482–500, April 2012.
Luke Jebaraj , Chakkaravarthy Venkatesan , Irisappane Soubache , and Charles Christober Asir Rajan .
Application of differential evolution algorithm in static and dynamic economic or emission dispatch problem: A
review. Renewable and Sustainable Energy Reviews, 77:1206–1220, 2017.
Brendan Jennings and Rolf Stadler . Resource management in clouds: Survey and research challenges.
Journal of Network and Systems Management, 23(3): 567–619, Jul 2015.
Yılmaz Kaya , Murat Uyar , Ramazan Tekđn . A novel crossover operator for genetic algorithms: Ring
crossover. 2011. https://arxiv.org/abs/1105.0355.
J. Kennedy and R. Eberhart . Particle swarm optimization. In Proceedings of ICNN'95 - International
Conference on Neural Networks, volume 4, pages 1942–1948, Nov 1995.
Paul C. Kocher . Timing Attacks on Implementations of Diffie-Hellman, RSA, DSS, and Other Systems. In Neal
Koblitz , editor, Advances in Cryptology– CRYPTO '96, pages 104–113, Berlin, Heidelberg, 1996. Springer
Berlin Heidelberg.
Shiann-Rong Kuang , Kun-Yi Wu , Bao-Chen Ke , Jia-Huei Yeh , and Hao-Yi Jheng . Efficient architecture and
hardware implementation of hybrid fuzzy-kalman filter for workload prediction. Integration, the VLSI Journal,
47(4): 408–416, 2014.
Jitendra Kumar , Rimsha Goomer , and Ashutosh Kumar Singh . Long short term memory recurrent neural
network (LSTM-RNN) based workload forecasting model for cloud datacenters. In Procedia Computer Science,
volume 125, pages 676–682. Elsevier B.V., 2018.
Jitendra Kumar , Deepika Saxena , Ashutosh Kumar Singh , and Anand Mohan . Biphase adaptive learning-
based neural network model for cloud datacenter workload forecasting. Soft Computing, 24:14593–14610, 10
2020.
Jitendra Kumar and Ashutosh Kumar Singh . Dynamic resource scaling in cloud using neural network and black
hole algorithm. In Proceedings on 5th International Conference on Eco-Friendly Computing and Communication
Systems, ICECCS 2016, pages 63–67. Institute of Electrical and Electronics Engineers Inc., 4 2017.
Jitendra Kumar and Ashutosh Kumar Singh . Workload prediction in cloud using artificial neural network and
adaptive differential evolution. Future Generation Computer Systems, 81:41–52, 4 2018.
Jitendra Kumar and Ashutosh Kumar Singh . Cloud resource demand prediction using differential evolution
based learning. In 2019 7th International Conference on Smart Computing and Communications, ICSCC 2019.
Institute of Electrical and Electronics Engineers Inc., 6 2019.
Jitendra Kumar and Ashutosh Kumar Singh . An efficient machine learning approach for virtual ma-chine
resource demand prediction. International Journal of Advanced Science Technology, 123:21–30, 2019.
Jitendra Kumar and Ashutosh Kumar Singh . Adaptive learning based prediction framework for cloud
datacenter networks' workload anticipation. Journal of Information Science and Engineering, 36:981–992, 2020.
Jitendra Kumar and Ashutosh Kumar Singh . Cloud datacenter workload estimation using error preventive time
series forecasting models. Cluster Computing, 23:1363–1379, 6 2020.
Jitendra Kumar and Ashutosh Kumar Singh . Decomposition based cloud resource demand prediction using
extreme learning machines. Journal of Network and Systems Management, 28:1775–1793, 10 2020.
Jitendra Kumar and Ashutosh Kumar Singh . Performance assessment of time series forecasting models for
cloud datacenter networks' workload prediction. Wireless Personal Communications, 116:1949–1969, 2 2021.
Jitendra Kumar , Ashutosh Kumar Singh , and Rajkumar Buyya . Ensemble learning based predictive
framework for virtual machine resource request prediction. Neurocomputing, 397:20–30, 7 2020.
Jitendra Kumar , Ashutosh Kumar Singh , and Rajkumar Buyya . Self directed learning based workload
forecasting model for cloud resource management. Information Sciences, 543:345–366, 1 2021.
Jitendra Kumar , Ashutosh Kumar Singh , and Anand Mohan . Resource-efficient load-balancing framework for
cloud data center networks. ETRI Journal, 43: 53–63, 2 2021.
Mohit Kumar and S. C. Sharma . Dynamic load balancing algorithm to minimize the makespan time and utilize
the resources effectively in cloud environment. International Journal of Computers and Applications, pages
1–10, Nov 2017.
P. Ravi Kumar , P. Herbert Raj , and P. Jelciana . Exploring Data Security Issues and Solutions in Cloud
Computing. Procedia Computer Science, 125:691–697, Jan 2018.
J. K. Lenstra and A. H. G. Rinnooy Kan . Complexity of vehicle routing and scheduling problems. Networks,
11(2):221–227, 2006.
L. Lerman and O. Markowitch . Efficient profiled attacks on masking schemes. IEEE Transactions on
Information Forensics and Security, 14(6):1445–1454, June 2019.
Xin Li , Zhuzhong Qian , Sanglu Lu , and Jie Wu . Energy efficient virtual machine placement algorithm with
balanced and improved resource utilization in a data center. Mathematical and Computer Modelling,
58(5):1222–1235, 2013.
Lingyun Yang , I. Foster , and J. M. Schopf . Homeostatic and tendency-based CPU load predictions. In
Proceedings International Parallel and Distributed Processing Symposium, pages 1–9. IEEE Comput. Soc, Apr
2003.
Ang Ee Mae , Wee Kuok Kwee , Pang Ying Han , and Lau Siong Hoe . Resource Allocation for Real-time
Multimedia Applications in LTE's Two-level Scheduling Framework. International Journal of Computer Science,
43(4):1–11, 2016.
Spyros Makridakis , Steven C. Wheelwright , and Rob J. Hyndman . Forecasting: Methods and Applications.
Wiley, 3 edition, 1 1998.
Valter Rogério Messias , Julio Cezar Estrella , Ricardo Ehlers , Marcos José Santana , Regina Carlucci
Santana , and Stephan Reiff-Marganiec . Combining time series prediction models using genetic algorithm to
autoscaling web applications hosted in the cloud infrastructure. Neural Computing and Applications,
27(8):2383–2406, Nov 2016.
E. Mezura-Montes and Carlos A. Coello Coello . An empirical study about the usefulness of evolution strategies
to solve constrained optimization problems. International Journal of General Systems, 37(4):443–473, 2008.
L. Minas and B. Ellison . Energy Efficiency for Information Technology: How to Reduce Power Consumption in
Servers and Data Centers. Intel Press, 2009.
Banaja Mohanty and Sasmita Tripathy . A teaching learning based optimization technique for optimal location
and size of DG in distribution network. Journal of Electrical Systems and Information Technology, 3(1):33–44,
2016.
Sathyan Munirathinam and B. Ramadoss . Big data predictive analtyics for proactive semiconductor equipment
maintenance. In 2014 IEEE International Conference on Big Data (Big Data), pages 893–902. IEEE, Oct 2014.
Sathyan Munirathinam and Balakrishnan Ramadoss . Predictive Models for Equipment Fault Detection in the
Semiconductor Manufacturing Process. International Journal of Engineering and Technology, 8(4):273–285,
Apr 2016.
San Murugesan . Cloud computing: A new paradigm in IT that has the power to transform emerging markets.
International Journal on Advances in ICT for Emerging Regions, 4(2):4–11, Oct 2008.
Nong Nurnie , Mohd Nistah , King Hann Lim , Lenin Gopal , Firas Basim , and Ismail Alnaimi . Coal-Fired Boiler
Fault Prediction using Artificial Neural Networks. International Journal of Electrical and Computer Engineering
(IJECE), 8(4):2486–2493, 2018.
Eva Patel and Dharmender Singh Kushwaha . Analysis of workloads for cloud infrastructure capacity planning.
In Lakhmi C. Jain , Valentina E. Balas , and Prashant Johri , editors, Data and Communication Networks, pages
29–42. Springer Singapore, Singapore, 2019.
G Pavai and TV Geetha . A survey on crossover operators. ACM Computing Surveys (CSUR), 49(4):72, 2017.
Satish Penmatsa , Gurdeep S Hura , and Princess Anne . Adaptive Cost Optimization and Fair Resource
Allocation in Computational Grid Systems. In 29th International Conference on Computer Applications in
Industry and Engineering (CAINE 2016), pages 1–6, Denver, Colorado, USA, 2016.
V. P. Plagianakos , D. K. Tasoulis , and M. N. Vrahatis . A Review of Major Application Areas of Differential
Evolution, pages 197–238. Springer Berlin Heidelberg, Berlin, Heidelberg, 2008.
J. J. Prevost , K. Nagothu , B. Kelley , and M. Jamshidi . Prediction of cloud data center networks loads using
stochastic and neural models. In 2011 6th International Conference on System of Systems Engineering, pages
276–281, June 2011.
Kenneth V Price. Differential evolution: a fast and simple numerical optimizer. In Fuzzy Information Processing
Society, 1996. NAFIPS., 1996 Biennial Conference of the North American, pages 524–527. IEEE, June 1996.
K.V. Price , R.M. Storn , and J.A. Lampinen . Differential Evolution: A Practical Approach to Global
Optimization. Natural Computing. Springer London, Limited, 2005.
A. K. Qin and P. N. Suganthan . Self-adaptive differential evolution algorithm for numerical optimization. In 2005
IEEE Congress on Evolutionary Computation, volume 2, pages 1785–1791, Sep. 2005.
Dang Minh Quan , Federico Mezza , Domenico Sannenli , and Raffaele Giafreda . T-alloc: A practical energy
efficient resource allocation algorithm for traditional data centers. Future Generation Computer Systems,
28(5):791–800, 2012. Special Section: Energy efficiency in large-scale distributed systems.
P Herbert Raj , P Ravi Kumar , and P Jelciana . Mobile Cloud Computing: A survey on Challenges and Issues.
International Journal of Computer Science and Information Security, 14(12):165–170, 2016.
R.V. Rao , V.J. Savsani , and D.P. Vakharia . Teaching-learning based optimization: A novel method for
constrained mechanical design optimization problems. Computer-Aided Design, 43(3):303–315, 2011.
Esmat Rashedi , Hossein Nezamabadi-Pour , and Saeid Saryazdi . GSA: A gravitational search algorithm.
Information Sciences, 179(13):2232–2248, 2009.
Charles Reiss , John Wilkes , and Joseph L. Hellerstein . Google cluster-usage traces: format + schema.
Technical report, Google Inc., Mountain View, CA, USA, November 2011.
I. Rodríguez-Fdez , A. Canosa , M. Mucientes , and A. Bugarín . STAC: A web platform for the comparison of
algorithms using statistical tests. In 2015 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE),
pages 1–8, Aug 2015.
Deepika Saxena and Ashutosh Kumar Singh . Auto-adaptive learning-based workload forecasting in dynamic
cloud environment. International Journal of Computers and Applications, 2020.
Deepika Saxena and Ashutosh Kumar Singh . Security embedded dynamic resource allocation model for cloud
data centre. Electronics Letters, 56:1062–1065, 9 2020.
Deepika Saxena and Ashutosh Kumar Singh . Energy aware resource efficient-(EARE) server consolidation
framework for cloud datacenter. In Lecture Notes in Electrical Engineering, volume 668, pages 1455–1464.
Springer, 2021.
Deepika Saxena and Ashutosh Kumar Singh . A proactive autoscaling and energy-efficient VM allocation
framework using online multi-resource neural network for cloud data center. Neurocomputing, 426:248–264, 2
2021.
Deepika Saxena , Ashutosh Kumar Singh , and Rajkumar Buyya . OP-MLB: An online VM prediction based
multi-objective load balancing framework for resource management at cloud datacenter. IEEE Transactions on
Cloud Computing, 2021.
Stefano Secci and San Murugesan . Cloud Networks: Enhancing Performance and Resiliency. Computer,
47(10):82–85, Oct 2014.
N. K. Sharma and G. R. M. Reddy . Multi-objective energy efficient virtual machines allocation at the cloud data
center. IEEE Transactions on Services Computing, 12(1):158–171, Jan 2019.
Vartika Sharma , Sizman Kaur , Jitendra Kumar , and Ashutosh Kumar Singh . A fast parkinson's disease
prediction technique using PCA and artificial neural network. In 2019 International Conference on Intelligent
Computing and Control Systems, ICCS 2019, pages 1491–1496. Institute of Electrical and Electronics
Engineers Inc., 5 2019.
Ashutosh Kumar Singh and Jitendra Kumar . Secure and energy aware load balancing framework for cloud
data centre networks. Electronics Letters, 55: 540–541, 2019.
Ashutosh Kumar Singh and Deepika Saxena . A cryptography and machine learning based authentication for
secure data-sharing in federated cloud services environment. Journal of Applied Security Research, 2021.
Ashutosh Kumar Singh , Deepika Saxena , Jitendra Kumar , and Vrinda Gupta . A quantum approach towards
the adaptive prediction of cloud workloads. IEEE Transactions on Parallel and Distributed Systems, pages 1–1,
2021.
Niharika Singh and Ashutosh Kumar Singh . Data privacy protection mechanisms in cloud. Data Science and
Engineering, 3:24–39, 3 2018.
Rainer Storn and Kenneth Price . Differential evolution: A simple and efficient heuristic for global optimization
over continuous spaces. J. of Global Optimization, 11(4):341–359, December 1997.
M. Tang , M. Luo , J. Zhou , Z. Yang , Z. Guo , F. Yan , and L. Liu . Side-channel attacks in a real scenario.
Tsinghua Science and Technology, 23(5):586–598, Oct 2018.
Ruey S. Tsay . Time series and forecasting: Brief history and future research. Journal of the American
Statistical Association, 95(450):638–643, 2000.
Manu Vardhan , Shrabani Mallick , Shakti Mishra , and D. S. Kushwaha . A Demand Based Load Balanced
Service Replication Model. Journal on Computing, 2(4):5–10, 2018.
Frank Wilcoxon . Individual comparisons by ranking methods. Biometrics Bulletin, 1(6):80–83, 1945.
Bo K Wong , Thomas A Bodnovich , and Yakup Selvi . Neural network applications in business: A review and
analysis of the literature (1988–1995). Decision Support Systems, 19(4):301–320, 1997.
Alden H Wright . Genetic algorithms for real parameter optimization. In Foundations of genetic algorithms,
volume 1, pages 205–218. Elsevier, 1991.
R. Xu , L. Zhu , A. Wang , X. Du , K. R. Choo , G. Zhang , and K. Gai . Side-channel attack on a protected RFID
card. IEEE Access, 6:58395–58404, 2018.
Preetesh K. Yadav , Sourav Pareek , Saif Shakeel , Jitendra Kumar , and Ashutosh Kumar Singh .
Advancements and security issues of IoT cyber physical systems. In 2019 International Conference on
Intelligent Computing and Control Systems, ICCS 2019, pages 940–945. Institute of Electrical and Electronics
Engineers Inc., 5 2019.
Xin-She Yang . Nature-Inspired Metaheuristic Algorithms. Luniver Press, 2008.
Xin She Yang and Xingshi He . Firefly algorithm: recent advances and applications. International Journal of
Swarm Intelligence, 1(1):36–50, 2013.
Q. Zhang , L. T. Yang , Z. Yan , Z. Chen , and P. Li . An efficient deep learning model to predict cloud workload
for industry informatics. IEEE Transactions on Industrial Informatics, 14(7):3170–3178, 2018.
Xinqian Zhang , Tingming Wu , Mingsong Chen , Tongquan Wei , Junlong Zhou , Shiyan Hu , and Rajkumar
Buyya . Energy-aware virtual machine allocation for cloud with resource reservation. Journal of Systems and
Software, 147:147–161, 2019.