A development approach for collective opportunistic Edge-of-Things services

Highlights

• Collective Opportunistic IoT Services are key abstractions for novel Internet- and Edge-of-Things applications.

• IoT Modelling may consider entire collectives and various infrastructural elements as first-class entities.

• Services implemented with Aggregate Computing can run in Edge or Cloudbased deployments, providing different pros and cons.

Abstract

Technological advances have recently fostered the Internet of Things vision, in which systems of situated entities perceive and act upon the world, and interact with one another to provide novel kinds of services, which are inherently cyber-physical, increasingly contextual and opportunistic in nature, and possibly span different scales and domains. The requirements of such IoT applications, however, pose significant non/functional challenges to engineering efforts, mitigated by emerging computing paradigms. On the infrastructure side, Cloud, Fog, and Edge Computing provide virtualised, on-demand, elastic resource provisioning – at the distant data centres, Network core and Edge – supporting the abstraction and scalability needs of IoT settings while also altogether giving options for QoS-driven trade-offs. However, despite intense research in these fields, there is still a gap of approaches supporting the engineering of dynamic, heterogeneous smart environments, such as those involving “collectives” of devices coordinating in a complex fashion to provide “global” services. In this paper, we integrate the Aggregate Computing and Opportunistic IoT Service models and propose a full-fledged approach for the engineering – from analysis to simulation – of complex “Edge of Things” applications. We compare by simulation two deployment targets for the same collective application: one centralised/Cloud-based, and the other decentralised/Edge-based. We discuss the trade-offs each one introduces, and we draw recommendations on application-driven choices of the appropriate deployment.

Introduction

Recent advances in Information and Communication Technology have fostered the Internet of Things (IoT) vision, where heterogeneous entities situated in our environments (i.e., cities, places, homes, and bodies) and capable of sensing, actuation, and processing, are interconnected with one another and with other infrastructural components in order to enable novel kinds of applications and services. Such new services tend to exploit what the environment and infrastructure have to offer, and are increasingly cyber-physical, context-aware, opportunistic and collective, with the staunch goal of unleashing the full Internet of Things ecosystem potential [16], [17], [18]. In other words, the environment itself is made smart through Smart Objects (SOs) and innovative services based on those which also span multiple scales and domains [30]—with examples including energy-optimised management of smart buildings [24], augmented logistics capabilities in smart ports [39], infrastructural maintenance in smart cities [13], energy-efficient monitoring in ecological scenarios [32], cognitive robot-supported smart manufacturing [22], etc.

However, the “layout” of smart environments is often based on heterogeneous infrastructure (e.g., access points, base stations) and heterogeneous devices with different physical forms, communication capabilities, energetic needs, storage capacities, and processing power. Concerning storage and processing, the Cloud Computing paradigm extends the capabilities of SOs by allowing them to offload data and computations to distant data centres. Moreover, the recent Fog [7] and Edge Computing [33] paradigms promise to provide the benefits of Cloud Computing without incurring in its problems (e.g., high latency, data rate consumption etc.), by bringing virtualised infrastructure and on-demand, elastic resource provisioning closer to end-devices, at the Edge or core of the network. Actually, Edge Computing is not an alternative to Cloud Computing but extends and complements it, by inclusively supporting resource-constrained devices and near real-time scenarios, and paving the path to Quality of Service (QoS) improvements. However, all this heterogeneity pushes additional complexity to engineering endeavours and stresses traditional approaches, methods and programming paradigms [15].

In this paper, we present an approach for seamlessly supporting the development (from the analysis to simulation) of collective, opportunistic Edge-of-Things (EoT) applications [14]. In particular, we model Smart Environments as ecosystems hosting opportunistic IoT services that may exploit Edge and cloud resources and may be collective in nature. The approach integrates the Opportunistic IoT Service model and the Aggregate Computing paradigm, whose conceptual alignment has been shown in [9]. As proof of concept, in a Smart City scenario, we show a crowd detection application which is first modelled as an Opportunistic IoT Service, then designed by following either a centralised/Cloud-based and a distributed/Edge-based approach, and finally implemented and simulated according to the Aggregate Computing paradigm. Simulation results show how a distributed/Edge-based approach can provide improved reactivity of collective applications at the expense of precision due to data from different SOs being processed in different Edge Servers.

The novel contributions with respect to our previous work [9] are threefold:

• A unified domain model for describing Collective and Opportunistic IoT Services to be provided both at the core and at the Edge of the network;

• A computation and communication performance model for enabling a quantitative analysis of the performed simulations;

• A detailed comparison and result analysis of the centralised/Cloud-based and distributed/Edge-based approaches, in order to analyse their pros and cons in supporting EoT applications.

The remainder of the paper is organised as follows. Section 2 provides a brief overview about the SO-based IoT vision – highlighting the importance of SOs in service provision and eliciting the features characterising an Opportunistic IoT Service – as well as about main concepts of emerging paradigms such as Cloud, Fog and Edge Computing. Such background is functional to introduce in Section 3 what we intend for EoT applications and our approach for their full-fledged development. This approach integrates Opportunistic IoT Service models and Aggregate Computing paradigm thus providing theoretical concepts and practical tools for engineering EoT applications from the modelling to the simulation phase. A crowd detection EoT application based on a real world urban mass event data set is developed according to the proposed approach in Section 4. We first describe it through the Opportunistic IoT Service model and then evaluate its performance in terms of responsiveness (measured as mean packet delay) and precision (measured as the number of SOs which are alerted of not entering crowded areas) in different scenarios of centralised/Cloud-based and decentralised/Edge-based computing. Finally, conclusions are drawn and future work delineated.