1
CLOUD ARCHITECTURES, NETWORKS, SERVICES, AND MANAGEMENT

Raouf Boutaba1 and Nelson L. S. da Fonseca2

1 D.R. Cheriton School of Computer Science, University of Waterloo, Waterloo, Ontario, Canada

2 Institute of Computing, State University of Campinas, Campinas, São Paulo, Brazil

1.1 INTRODUCTION

With the wide availability of high-bandwidth, low-latency network connectivity, the Internet has enabled the delivery of rich services such as social networking, content delivery, and e-commerce at unprecedented scales. This technological trend has led to the development of cloud computing, a paradigm that harnesses the massive capacities of data centers to support the delivery of online services in a cost-effective manner. In a cloud computing environment, the traditional role of service providers is divided into two: cloud providers who own the physical data center and lease resources (e.g., virtual machines or VMs) to service providers; and service providers who use resources leased by cloud providers to execute applications. By leveraging the economies-of-scale of data centers, cloud computing can provide significant reduction in operational expenditure. At the same time, it also supports new applications such as big-data analytics (e.g., MapReduce [1]) that process massive volumes of data in a scalable and efficient fashion. The rise of cloud computing has made a profound impact on the development of the IT industry in recent years. While large companies like Google, Amazon, Facebook, and Microsoft have developed their own cloud platforms and technologies, many small companies are also embracing cloud computing by leveraging open-source software and deploying services in public clouds.

However, despite the wide adoption of cloud computing in the industry, the current cloud technologies are still far from unleashing their full potential. In fact, cloud computing was known as a buzzword for several years, and many IT companies were uncertain about how to make successful investment in cloud computing. Fortunately, with the significant attraction from both industry and academia, cloud computing is evolving rapidly, with advancements in almost all aspects, ranging from data center architectural design, scheduling and resource management, server and network virtualization, data storage, programming frameworks, energy management, pricing, service connectivity to security, and privacy.

The goal of this chapter is to provide a general introduction to cloud networking, services, and management. We first provide an overview of cloud computing, describing its key driving forces, characteristics and enabling technologies. Then, we focus on the different characteristics of cloud computing systems and key research challenges that are covered in the subsequent 14 chapters of this book. Specifically, the chapters delve into several topics related to cloud services, networking and management including virtualization and software-defined network technologies, intra- and inter- data center network architectures, resource, performance and energy management in the cloud, survivability, fault tolerance and security, mobile cloud computing, and cloud applications notably big data, scientific, and multimedia applications.

1.2 PART I: INTRODUCTION TO CLOUD COMPUTING

1.2.1 What Is Cloud Computing?

Despite being widely used in different contexts, a precise definition of cloud computing is rather elusive. In the past, there were dozens of attempts trying to provide an accurate yet concise definition of cloud computing [2]. However, most of the proposed definitions only focus on particular aspects of cloud computing, such as the business model and technology (e.g., virtualization) used in cloud environments. Due to lack of consensus on how to define cloud computing, for years cloud computing was considered a buzz word or a marketing hype in order to get businesses to invest more in their IT infrastructures. The National Institute of Standards and Technology (NIST) provided a relatively standard and widely accepted definition of cloud computing as follows: "cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction." [3]

NIST further defined five essential characteristics, three service models, and four deployment models, for cloud computing. The five essential characteristics include the following:

1. On-demand self-service, which states that a consumer (e.g., a service provider) can acquire resources based on service demand;
2. Broad network access, which states that cloud services can be accessed remotely from heterogeneous client platforms (e.g., mobile phones);
3. Resource pooling, where resources are pooled and shared by consumers in a multi-tenant fashion;
4. Rapid elasticity, which states that cloud resources can be rapidly provisioned and released with minimal human involvement;
5. Measured service, which states that resources are controlled (and possibly priced) by leveraging a metering capability (e.g., pay-per-use) that is appropriate to the type of the service.

These characteristics provide a relatively accurate picture of what cloud computing systems should look like. It should be mentioned that not every cloud computing system exhibits all five characteristics listed earlier. For example, in a private cloud, where the service provider owns the physical data center, the metering capability may not be necessary because there is no need to limit resource usage of the service unless it is reaching data center capacity limits. However, despite the definition and aforementioned characteristics, cloud computing can still be realized in a large number of ways, and hence one may argue the definition is still not precise enough. Today, cloud computing commonly refers to a computing model where services are hosted using resources in data centers and delivered to end users over the Internet. In our opinion, since cloud computing technologies are still evolving, finding the precise definition of cloud computing at the current moment may not be the right approach. Perhaps once the technologies have reached maturity, the true definition will naturally emerge.

1.2.2 Why Cloud Computing?

In this section, we present the motivation behind the development of cloud computing. We will also compare cloud computing with other parallel and distributed computing models and highlight their differences.

1.2.2.1 Key Driving Forces.

There are several driving forces behind the success of cloud computing. The increasing demand for large-scale computation and big data analytics and economics are the most important ones. But other factors such as easy access to computation and storage, flexibility in resource allocations, and scalability play important roles.

Large-scale computation and big data: Recent years have witnessed the rise of Internet-scale applications. These applications range from social networks (e.g., facebook, twitter), video applications (e.g., Netflix, youtube), enterprise applications (e.g., SalesForce, Microsoft CRM) to personal applications (e.g., iCloud, Dropbox). These applications are commonly accessed by large numbers of users over the Internet. They are extremely large scale and resource intensive. Furthermore, they often have high performance requirements such as response time. Supporting these applications requires extremely large-scale infrastructures. For instance, Google has hundreds of compute clusters deployed worldwide with hundreds of thousands of servers. Another salient characteristic is that these applications also require access to huge volumes of data. For instance, Facebook stores tens of petabytes of data and processes over a hundred terabytes per day. Scientific applications (e.g., brain image processing, astrophysics, ocean monitoring, and DNA analysis) are more and more deployed in the cloud. Cloud computing emerged in this context as a computing model designed for running large applications in a scalable and cost-efficient manner by harnessing massive resource capacities in data centers and by sharing the data center resources among applications in an on-demand fashion.

Economics: To support large-scale computation, cloud providers rely on inexpensive commodity hardware offering better scalability and performance/price ratio than supercomputers. By deploying a very large number of commodity machines, they leverage economies of scale bringing per unit cost down and allowing for incremental growth. On the other hand, cloud customers such as small and medium enterprises, which outsource their IT infrastructure to the cloud, avoid upfront infrastructure investment cost and instead benefit from a pay-as-you-go pricing and billing model. They can deploy their services in the cloud and make them quickly available to their own customers resulting in short time to market. They can start small and scale up and down their infrastructure based on their customers demand and pay based on usage.

Scalability: By harnessing huge computing and storage capabilities, cloud computing gives customers the illusion of infinite resources on demand. Customers can start small and scale up and down resources as needed.

Flexibility: Cloud computing is highly flexible. It allows customers to specify their resource requirements in terms of CPU cores, memory, storage, and networking capabilities. Customers are also offered the...