Page 44 - European Energy Innovation - Spring 2016 publication
P. 44

44  Spring 2016 European Energy Innovation


Cognitive computation and
communication for IoT

By Van Tam Nguyen, UC Berkeley / Telecom ParisTech

1. INTRODUCTION                              to external attacks that can cause either  expenditure and emission of CO2.
In the next decades, Internet of Things      leak of private information or dangers     Today, the DCs are already responsible
(IoT), the interconnected networks           to users. Third, many IoT applications     for about 2% of global greenhouse gas
of physical objects embedded with            require a high degree of reliability,      emissions, a similar share to aviation4.
electronics, software, sensors, and          robustness and a low degree of latency     In 2007, the DCs consumed on the
connectivity will revolutionize how we       that exceeds the current design of         order of 330bn kWh, equivalent to the
work, live, exercise, entertain and travel.  wireless communication and cloud           entire electricity demand of the UK.
IoT is experiencing explosive growth in      computing. Fourth, since IoT requires      This demand is projected to triple or
both quantity (20.8 billion IoT devices      process of large amount of data from       quadruple by 2020, and accounts for
by 20201) and utility, with increasingly     numerous devices, hardware design          1.5-2% of all global electricity demand,
important applications in healthcare,        for IoT applications need to be not only   at a growing rate of 12% per year5.
military operations, transportation and      flexible and adaptive but also highly      Many IoT applications, such as smart
urban planning2. However, IoT faces          energy-efficient.                          vehicular traffic management system,
several major growing challenges. First,                                                smart driving and smart grid require
incorporating appropriate intelligence       In the current architecture of IoT,        real-time and low-latency services. If the
and smart connectivity into IoT objects      cloud computing provides the virtual       processing, computation and storage
requires a computing paradigm                infrastructure for data collection,        of the enormous amount of data are
that exceeds the current computing           analysis, visualization and service        performed only within DCs, the massive
capabilities of smart phones and             delivery³. With the growing number of      data traffic generated from IoT devices
portables3. Second, ensuring privacy,        billions of IoT devices, there will be a   will result in network bottlenecks,
security and safety of IoT applications is   great demand on cloud Data Centres         and affect the performance of all IoT
critically important, as IoT is susceptible  (DCs), resulting in massive energy         applications. In order to better handle
                                                                                        the communication demand of the IoT,
                                                                                        and reduce the energy consumption
                                                                                        and the emission of CO2, Bonomi
                                                                                        et al. proposed the concept of Fog
                                                                                        computing6. Its key principle is to
                                                                                        bring the cloud closer to the end user
                                                                                        by transforming as much as possible
                                                                                        data into action at the network edge.
                                                                                        The recent work in7 showed that in the
                                                                                        context of high number of latency-
                                                                                        sensitive applications, Fog computing
                                                                                        outperforms cloud computing.
                                                                                        However, the privacy issues, the
                                                                                        security and reliability problems remain

Fig. 1: COGNICOM concept                                                                2 COGNICOM SOLUTION
                                                                                        To address those four major
                                                                                        challenges, we propose the
                                                                                        development of COGNICOM, a hybrid
                                                                                        architecture powered by Cognitive
                                                                                        Engine (CE) that facilitates optimal

   39   40   41   42   43   44   45   46   47   48   49