Machine Learning for Cloud Management

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

• Éditeur: Chapman and Hall/CRC
• Année: 2021
• ISBN:
  9781000476613
• Édition: 1
• Sujet: Informatica
• Sous-sujet: Informatica generale

Cloud computing offers subscription-based on-demand services, and it has emerged as the backbone of the computing industry. It has enabled us to share resources among multiple users through virtualization, which creates a virtual instance of a computer system running in an abstracted hardware layer. Unlike early distributed computing models, it offers virtually 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, a number of users, and the amount of hosted data. The large and complex workloads hosted on these data centers introduce many challenges, including resource utilization, power consumption, scalability, and operational cost. Therefore, an effective resource management scheme is essential to achieve operational efficiency with improved elasticity. Machine learning enabled solutions are the best fit to address these issues as they can analyze and learn from the data. Moreover, it brings automation to the solutions, which is an essential factor in dealing with large distributed systems in the cloud paradigm.

Machine Learning for Cloud Management explores cloud resource management through predictive modelling and virtual machine placement. The predictive approaches are developed using regression-based time series analysis and neural network models. The neural network-based models are primarily trained using evolutionary algorithms, and efficient virtual machine placement schemes are developed using multi-objective genetic algorithms.

Key Features:

• The first book to set out a range of machine learning methods for efficient resource management in a large distributed network of clouds.
• Predictive analytics is an integral part of efficient cloud resource management, and this book gives a future research direction to researchers in this domain.
• It is written by leading international researchers.

The book is ideal for researchers who are working in the domain of cloud computing.

CHAPTER 1 Introduction

DOI: 10.1201/9781003110101-1

Cloud computing paradigm enables the delivery of computing resources and applications to users across the globe as subscription-oriented services. Virtualization is the technique behind the scene that helps in resource sharing among multiple users in this cloud computing environment.

1.1Cloud 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 services 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.

Figure 1.1Service model view of cloud computing

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 [63, 62]. Modern cloud systems are equipped with characteristics such as on-demand service, reliability, scalability, elasticity, disaster recovery, accessibility, measured services, and many others [98, 73, 75, 25]. However, various challenges and limitations are still open including inefficient resource management, security and privacy, heterogeneity, elasticity, usability, response time, and many more [21, 18, 109, 90, 124, 19, 125, 53, 54, 52].

1.2Cloud Management

Resource management is one of the core functions of cloud systems and must be improved for better system performance [102, 119, 66]. 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, 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...

Table des matières

• Cover Page
• Half-Title Page
• Title Page
• Copyright Page
• Dedication Page
• Contents
• List of Figures
• List of Tables
• Preface
• Author
• Abbreviations
• CHAPTER 1 ◾ Introduction
• CHAPTER 2 ◾ Time Series Models
• CHAPTER 3 ◾ Error Preventive Time Series Models
• CHAPTER 4 ◾ Metaheuristic Optimization Algorithms
• CHAPTER 5 ◾ Evolutionary Neural Networks
• CHAPTER 6 ◾ Self Directed Learning
• CHAPTER 7 ◾ Ensemble Learning
• CHAPTER 8 ◾ Load Balancing
• CHAPTER 9 ◾ Summary
• Bibliography
• Index