Cloud computing, defined by the National Institute of Standards and Technology (NIST) as “a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources” [1], offers appealing benefits when addressing the challenges of today’s data-centered world in terms of availability and scalability, where self-service provisioning and elasticity can enable flexible adaptation to variations in workloads.
Many important developments in this field were led by players such as Amazon, eBay, Facebook, and Google, to build and run their proprietary architectures, but now, with the maturation of the market for infrastructure as a service (IaaS) and platform as a service (PaaS), businesses, no matter how small, are able to take advantage of cloud offerings to access otherwise unachievable capabilities in a cost-affordable way.
Still, access to enabling computing platforms does not represent the definitive solution for the successful deployment of data-intensive cloud applications. In fact, careful consideration has to be given to issues such as availability, confidentiality, data transfer performance, and application parallelization, as well as to risks of vendor lock-in. From this perspective, the book appears particularly relevant as it offers timely coverage of the requirements, characteristics, and challenges of managing large-scale cloud-based data services.
Chapters 1 and 2 offer an introduction to cloud computing concepts and provide a depiction of the state of the art in public cloud platforms, while chapter 3 introduces cloud-based data storage systems with an emphasis on NoSQL database systems.
In chapter 4, the authors present the cloud architecture runtime evaluation (CARE) framework, which was developed to benchmark the performance of cloud platforms such as Amazon Web Services (AWS), Google App Engine, and Microsoft Azure, stress-testing aspects such as availability and reliability. One key finding is the relationship between the quality of the end-user network access and overall user experience. The specific aspects of NoSQL as a service are covered in chapter 5, which looks into database replication across different geographic regions, for example, to identify tradeoffs between high availability/low latency and strong consistency.
Chapter 6 expands the study of replicated databases beyond SQL and NoSQL by focusing on virtualized database servers, that is, the exploitation of virtualization to port existing data center applications to the cloud. Questions being looked at relate to the scalability of master-slave replication strategies and to the modeling of the average replication delay with different cardinalities and geographical distributions.
A major practical contribution of this book lies in chapter 7, which presents a service-level agreement (SLA)-driven framework for managing database replication, where typical SLAs are not designed to closely match the flexibility needs of customers. The framework proposed by the authors is specifically relevant because it is independent from the database platform of choice and does not require changes to the cloud-hosted software applications.
The final chapters of the book include a proposal for cloud service composition based on genetic algorithms to compute quality of service (QoS) targets, as well as an overview of modern approaches for large-scale data processing and related research challenges.
Overall, this is a very interesting piece of work, filled with a lot of timely information and written from a very practical perspective to include material on implementation costs for different data management options. A good and enjoyable read for advanced students and practitioners working in the field.
