IoT-Based Sensor Networks: Architectural Organization, Virtualization and Network Re-orchestration
MetadataShow full metadata
Internet of ‘Things’ (IoT), an extension of localised ‘Wireless’ Sensor Networks (WSN), has been employed to realize a multitude of smart, intelligent and pervasive Cyber Physical System (CPS) infrastructures. CPS encompasses a host of technological and architectural challenges like low-power communication, protocol conversions, data transport and the ability to interoperate with other IoT technologies. This makes CPS significantly complex and reduces its flexibility to adapt. A typical IoT-based sensor network may to a certain extent, lack key softwarization-enabled operational drivers that may introduce significant constraints on its ability to flexibly engage with its external surroundings. Flexible re-orchestration of such complex IoT based sensor networks, however, is vital towards aligning system ‘dynamics’ with that of a monitored ‘physical’ phenomenon while operating in dynamic physical environments (for example WSNs deployed for outdoor applications such as forest fire monitoring). Arguably, considerable levels of operational flexibility can be achieved through a cloud-based architectural framework that hosts the required operating tools to allow for software-defined network virtualization that enable suitable re-orchestrations of the related physical network. In a nutshell, research work documented within this thesis endeavours towards rendering a sensor network capable of undergoing desired flexible re-orchestrations via converging upon a novel architectural proposition inclusive of modularization, cloud-based virtualization, software control via ‘command-driven reconfigurability’, and maintenance of a library for additional firmware modules (for each of the nodes at the physical level), among others. Other equally noteworthy and innovative contributions of this thesis pertain to outlining of a seemingly logical strategy for the sensor network ‘re-orchestration’ process (that spans across ‘three’ phases of, ‘Data Analysis and Event-Identification’, ‘Re-orchestration-Planning’ and ‘Re-orchestration-Execution’) as well as both determining and formulating a generic model for the latency associated with the same. The approach adopted herein is to allow for the underlying physical layer to undergo desired node and network-level (including topological) re-orchestrations (based on the outcomes derived from the cloud) in a flexible and expeditious manner during run-time through a ‘Command-driven’ re-configurability approach. This relatively simplistic yet expedient approach involves loading of a ‘unified firmware’ (i.e., one encompassing the requisite, ‘well-defined’ software modules) onto nodes (assumed to be capable of accommodating for and executing the corresponding functional roles owing to the enhanced capabilities ushered in by the advancements attained in the field of SoC and Embedded Systems technologies) to allow for conditional execution of the same remotely by means of ‘commands’. In order to augment the flexibilities that could be offloaded by the node over time based on the service requirements, a library of ‘reusable firmware modules’ (within which the requisite new functional modules could be integrated from time-to-time) could be maintained to be readily accessible by the main firmware. In regard to the above context, it is deemed worthy to reiterate that the thesis underscores the key prerequisites for the above prior to laying the concept in chapter four. Firstly, this includes identifying and clearly defining the core functional components (constituting any IoT-based sensor network organization viz., ‘leaf’, router and ‘Gateway’ functionalities) as ‘modules. The second prerequisite pertains to modularization of the core functional components that have been identified and defined. Virtualization of the core functional modules so identified and thereby the entire network (essentially, cloud-level Network Virtualization i.e., ‘NV’) that ‘logically’ (i.e., from a software standpoint) mimics the operational dynamics of the underlying physical network functions will form the third prerequisite. As alluded to earlier, the fourth prerequisite refers to the library of reusable ‘firmware modules’ at the node level (for augmented flexibility). The thesis is sectioned into seven different chapters, each accounting for a specific element of the overall work. The first chapter provides an overview of the various technological domains and aspects associated with this research work, whilst laying out the necessary background, vision and motivation behind the same. The second chapter accounts for a review of the existing literature pertaining to the various elements associated with this research viz., WSN virtualization, softwarization, re-orchestration and associated network downtime (as well as other architectural frameworks designed with relatively similar motives in mind). Information pertaining to the tools employed for virtualization and hardware implementation purposes are provided in the third chapter. As elaborated above, Chapter 4 firstly spells out the key perquisites for the proposed architecture prior to describing the same, along with its internal components. It then outlines the strategy adopted for the re-orchestration process, including formulation of a generic model for the latency that the network may experience as a result of the same. By means of certain pertinent example cases of software-defined sensor network re-orchestrations, chapter 5 details the specifics of both virtual and physical implementations, conducted via utilizing the Contiki-oriented virtual platform of the Cooja simulator as well as the Contiki-ported Texas Instruments CC2538 wireless transceivers respectively. It also brings to the fore the practicability of employing Contiki as a tool for software development that allows for precise replication of the codes employed for physical motes at the virtual level, whilst leveraging on the same to better analyse and conduct more accurate performance evaluations pertaining to the re-orchestration process. As a means to demonstrate the workability of the proposed concept with respect to a real-life scenario, chapter 6 deals with the use case pertaining to forest fire monitoring wherein dynamic re-orchestration of sensor network so deployed could significantly aid (pre-emptive) re-routing of network dataflow and/or maintenance of network connectivity in the event of network fragmentation emanating out of rapidly spreading uncontained fire outbreaks. Chapter 7 puts forth the conclusion of this thesis work, along with the future course of work to be undertaken.