Скачиваний:
50
Добавлен:
20.06.2019
Размер:
50.48 Mб
Скачать

Preface

Introduction

Cloud computing appears to have emerged very recently as a subject of substantial industrial and academic interest, though its meaning, scope and fit with respect to other paradigms is hotly debated. For some researchers, Clouds are a natural evolu- tion towards full commercialisation of Grid systems, while for others they may be dismissed as a mere rebranding of the existing pay-per-use or pay-as-you-go technologies. From either perspective, it appears that ‘Cloud’ has become the label of choice for accountable pay-per-use access to a wide variety of third-party applications and computational resources on a massive scale. Clouds are now supporting patterns of less-predictable resource use for applications and services across the IT spectrum, from online office applications to high-throughput transactional services and high-performance computations involving substantial quantities of processing cycles and storage. The current notion of Clouds seems to blur the distinctions between Grid Services, Web Services, and data centres, amongst others, and brings considerations of lowering the cost for relatively bursty applications to the fore.

Currently, there appears to be an increasing demand for Cloud computing in general. Major IT and e-commerce vendors such as Amazon, Google, IBM,

Microsoft, and Sun have joined a variety of technology and service providers in offering Clouds. In turn, this generates significant demand for reference materials that provide coverage for this topic, ranging from standard developer guides to advanced expositions of research into Cloud design, optimisation and management. Interest in Cloud computing, as a concept or system design abstraction, is compounded and further strengthened by an inherent relationship to service-oriented computing. Clouds may be considered by some as a reincarnation and an extension of service-oriented computing that covers computational hardware-based resources as well as software, with concomitant business benefits in cost reduction where such services scale efficiently.

For the scientific community, Cloud computing offers interesting characteristics and challenges. Some of these exist at the intersection between computing and economics, where the key question is how to develop a Cloud infrastructure that provides the required quality of service; from a network economics perspective,

ix

x

Preface

this relates to the design and deployment of an adaptive pricing mechanism that provides both for a competitive edge and a profitable venture. At the same time, end-users (both potential consumers and providers) need to be able to understand the similarities, differences, benefits and disadvantages of Clouds over numerous existing paradigms including Grids, High-Performance Computing, Peer-to-Peer (P2P) systems and so on. P2P, Grid, High-Performance Computing and Web

Services are very pertinent fields that have received significant and sustained research interest in the design and deployment of large-scale and high-performance computational resource-sharing systems. Collectively, these form the de facto basis for methods and techniques that will be re-appraised, re-used or re-designed to construct performance-driven Cloud platforms capable of satisfying the four cor- nerstones of quality of service:

1.Efficiency: The execution and coordination of the services is optimised in terms of data traffic and latency. Data traffic is typically one of the main cost factors in any distributed computing framework and thus its reduction is a standard long- term goal of such systems. Latency is arguably one of the most important factors affecting customer satisfaction and therefore it should also be within specified acceptable limits.

2.Scalability: These platforms should scale well to massive customer bases. They must also withstand demand of multiple bursty applications during peak times and endure the ‘flash crowds’ phenomenon familiar in overly successful marketing strategies and provisioning for popular websites at key times.

3.Robustness: The services need continuously high availability by design, with effective use of redundancy and graceful failover. Where users are charged for the expected successful use of computational facilities, it is imperative to understand the risk of failure, either to remove the probability of failure, or to use this information to offer appropriate compensation schemes.

4.Security:Appropriatesecurityprovisionsmustexistforbothdataandapplicationsto protect both the providers and consumers from malicious or fraudulent activities. Without adequate security provisioning, it is highly unlikely that any commoditised platform would become a serious consideration for business computing.

To ensure commercial success, effective Clouds will be expected to provide guaranteed quality of service to customers by satisfying these four cornerstones.

This book is targeted at providing a thorough and advanced treatment of the state-of-the-art in Cloud computing that addresses the above topics and highlights and clarifies the conceptual and systemic links with other distributed computing approaches.

The book has four key objectives:

(i)To explore the relationship of Cloud computing to other distributed computing paradigms, namely Peer-to-Peer, Grids, High-Performance Computing and

Web Services

(ii)To present the principles, techniques, protocols and algorithms that can be adapted from other distributed computing paradigms to the development of successful Clouds

Preface

xi

(iii)to present current Cloud applications and highlight early deployment experiences

(iv)to elaborate the economic schemes needed for Clouds to become viable business models.

The first two objectives are firmly rooted in extant discourse of distributed computing and a desire to understand the potential of all these technologies in constructing purpose-specific hybrid solutions. The remaining objectives are closely linked to commercial demand for understanding how such technologies can shape successful and profitable businesses.

Expected Audience

This book should be of particular interest for the following audiences:

Researchers and doctoral students working specifically in Cloud computing research, implementation and deployment, primarily as a reference publication. Similarly, this book should be useful to researchers in related, or more general fields, such as distributed computing, software engineering, Web Services, modelling of business processes, and so on.

Academics and students engaging in research-informed teaching in the above fields. This book can serve as a good collection of articles to facilitate a broad understanding of this subject and as such may be useful as a key reference text in such teaching.

Professional system architects and developers who could decide to adapt and apply in practice a number of the techniques and processes presented in the book.

Technical managers and IT consultants as a book that demonstrates the potential applicability of certain methods for delivering efficient and secure commercial electronic services to customers globally.

These audiences will find this publication appealing as it combines three distinct scholarly contributions: first, it identifies and highlights state-of-the-art techniques and methods for designing Cloud systems; second, it presents mechanisms and schemes for linking Clouds to economic activities; third, it achieves balanced coverage of all related technologies that collectively contribute towards the realisation of Cloud computing.

Book Overview

The book contains 21 chapters that were carefully selected based on peer review by at least two expert and independent reviewers. The chapters are split into four Parts:

xii

Preface

Part 1: Cloud Base

This section aims to cover the essential definitions, characteristics and concepts behind Cloud computing. The chapters included in this section collectively introduce the reader to Cloud computing and its essential architectural principles. As a result, chapters in this section are either tutorial in nature or provide critical literature surveys in the field.

Chapter 1 presents a number of mainstream technologies for building and managing Cloud architectures. The authors provide a detailed description of virtual machine frameworks and present the MapReduce programming model, which is suitable for large-scale data processing.

Chapter 2 describes a detailed taxonomy of Cloud computing architectures that may promote clarity and reusability of key concepts in Cloud design. The authors use this taxonomy to identify key similarities and differences between various approaches to Cloud computing and underline areas for further development.

Chapter 3 analyses the applicability of Cloud computing in e-Science. It focuses on classifying different Cloud architectures in terms of their ability to provide the services required for large-scale scientific experiments and calculations.

Chapter 4 examines the differences between Cloud and Grid computing. The authors explain how the userand task-centric design philosophy of Clouds makes this technology more appealing to typical end-users.

Chapter 5 provides a high-level overview of various standards for, and related to, Cloud computing. It explores the key features of each standard in terms of interoperability, security and portability and assesses the potential for market adoption of the standards presented.

Part 2: Cloud Seeding

This section builds on the introductory material of Part 1 and provides in-depth coverage of how Clouds can be designed and how emerging technologies such as P2P fit with Cloud computing in general. It includes chapters that propose novel techniques and systems for making Clouds scalable, efficient and fault-tolerant computing platforms.

Chapter 6 presents an innovative computational paradigm called Cloud@Home that merges Peer-to-Peer computing with Clouds. The Cloud@Home aggregates the computational resources of many low-power systems, and the authors demonstrate how this pool of resources can subsequently be managed and used by different communities of users.

Chapter 7 exploits a novel peer-to-peer model for replicating and managing job states in an efficient and decentralised way. The authors use this model to enhance the fault tolerance of the MapReduce programming paradigm in highly dynamic environments that exhibit significant failure rates.

Соседние файлы в папке CLOUD COMPUTING