Browse Books

Reviewer: Damien Magoni

With the advent of the Internet of Things (IoT), many networking architectural choices made during recent decades have become ill adapted to the current evolution of the network of machines (that is, sensors, actuators, and so on). A new architecture is currently emerging, and this book is an attempt to describe what it would be like and how it should behave. The author has made efforts to present the content from a high-level point of view, and the book is thus appropriate for any audience, from undergrads to senior executives. Its structure is easy to follow and its content only requires a minimal background in technology. Published by Apress as part of the “Expert's Voice in Internet Technologies” series, it is under 200 pages long and provides material that can be used in a teaching context for delivering a broad introduction to the IoT. Eight chapters are included, which progress from the problems encountered to the description of an efficient architecture, and finally to the possible pathways for making it real. Chapter 1 presents the landscape of the IoT, especially its size. With dozens or even hundreds of billions of devices, the IoT will need to be divided into three functional classes: the end devices, the propagators, and the integrators. End devices are expected to be cheap dummy sensor devices mainly able to send some bits of data to propagators at periodic times. Propagators will be autonomous self-* devices able to aggregate and forward the data to integrators. Some of those will be gateways between the sensors' communication protocols and the Internet protocol (IP)-based Internet. Finally, integrators will serve as data analyzers and will be used by humans as interfaces to the IoT world. As such, they will also send orders to end devices when needed. Chapters 2 and 3 introduce a new communication protocol for the IoT. As IP will be too costly and complex to implement in end devices and as it would lead to huge overheads, the author proposes a new protocol exchanging data chunks defined as chirps. These packets are defined with variable-length header fields separated by markers, and they are self-classified. This means that a chirp header will contain a label defining the data it carries as opposed to a specific destination. Thus, the communication model is multicast-like, with destination addresses being logical topic names. Routing will also be easy, as only two types of paths will need to exist: from end devices to integrators and vice versa. The data selectivity will be receiver oriented, as in the publish/subscribe model. The new protocol is envisioned to be as simple as possible. It will sit on top of layer 2 protocols. If existing layer 2 protocols are too expensive, new ones could also be defined as needed. Reliability will be provided by the simple redundancy found in the massive amount of data emitted. The loss of a single packet will be irrelevant, as information will come out of the aggregation of big amounts of data. In chapters 4 and 5, the author deals with scalability and integration. The data will flow from many sensors to a few integrators via propagators. This will lead to a tree-shaped data flow that will inherently be scalable, as data will be aggregated along the way. To avoid breaking a path due to the failure of an intermediate node, changing the tree to a mesh will be possible to increase reliability. Two schemes will enable efficient bandwidth usage: dumping duplicates and aggregating different chirps inside one IP packet. A publishing agent controlled by integrators will also be present inside some propagator nodes to create and modify available data flows. Those flows will be made meaningful altogether by using neighborhood affinities. Some will be open to the public while others will be private and undecipherable by third parties. A key feature ensuring the scalability of the architecture is also the definition of one high (propagator-integrator) and one low (sensor-propagator) level loop, thus making up a dual isochronous control loop. Indeed having only one end-to-end control loop, like in regular IP networks, would be impossible due to the lack of determinism found at the edge of the IoT. Finally, as large amounts of data may arrive in an integrator, a filter gateway will be put at its front end to ensure that only meaningful and significant data is analyzed. Chapter 6 describes in more detail all of the concepts that were presented in the previous chapters. Some examples and applications are presented in chapter 7, while chapter 8 provides some so-called pathways to the realization of this architecture. Among these pathways, the use of open-source solutions as well as the need for proper flexible standards that leave enough room for expansion are emphasized. The book unfortunately contains no references to any related research work, although this would have been very handy to deepen the study of some topics. Apart from the low-level loop definition, no specific consideration has been given to the real-time requirements of most of the applications found in the IoT. Finally, there are also some repetitions through the chapters, but they are intended to emphasize the concepts. Overall, despite sometimes lacking precision on some specific issues (addressing the end devices, routing between the propagators, and so on), this book is very interesting and provides a lot of insights into the design of the IoT. It does fulfill its objectives in describing the IoT and the fundamental concepts that should be implemented in order to enable its success. More reviews about this item: Amazon Online Computing Reviews Service