Strategies to Implement Edge Computing in a P2P Pervasive Grid

Introduction

Big data and data analytics have become essential tools for the strategic planning of any company. While data analysis is not a new topic, it was boosted by the development of large-scale computing platforms, notably the clouds. While cloud computing relies on distant resources, several works try to leverage the use of proximity resources as effective computing platforms (Garcia Lopez, 2015; Parashar & Pierson, 2010; Steffenel & Kirsch-Pinheiro, 2015).

Indeed, the number and nature of proximity computing devices (smartphones, Internet of Things - IoT, etc.) is growing exponentially, and it is important to understand how to exploit the power of these computing resources. For this reason, new approaches like edge or fog computing aim at delegating part of the work to devices in the “edge” of the network (Lopez, 2015). Because several strategies can be used to implement edge computing, this work specifically focus on the use of pervasive grids to leverage such challenges. Indeed, pervasive grids (Parashar & Pierson, 2010) associate classical and volatile computing resources. We believe that organizations can perform big data analytics with minimal costs by associating IoT and mobile devices as well as idle or unused resources in the enterprise network.

Therefore, in this paper, we discuss the limitations of traditional big data computing platforms and introduce the principles and challenges to implement edge computing over pervasive grids. To illustrate this, we present CloudFIT, a distributed computing platform based on a P2P overlay, and discuss how it can be improved to efficiently deploy edge/pervasive computing applications, especially those related to big data analytics. Indeed, after presenting the main architecture of CloudFIT, we focus on the required strategies to ensure data proximity and context awareness, two factors that impact the performance of big data analysis in distributed systems.

For conducting this research, we adopted a two-fold approach, combining a conceptual research method with a case study a case study. Indeed, according to research method categories pointed out by Mora et al. (2008), a conceptual research corresponds to the study of ideas related to real objects including designing of new conceptual artifacts such as a framework/model, a method/model, or a system/component. For these authors, a “conceptual design research is the purposeful design of conceptual artifacts”, in which the design artifact is dictated by the design goals. The principles and challenges we discuss in this paper represent, in our research, these design goals that guided the application of CloudFIT platform. The results of this conceptual research are then confronted to a case study issue from a real geophysical problem. Peres, 2013 conduct a case study analysis of the detection of Ozone Secondary Events (OSE) problem and Peres et al., (2017) present a detailed description of TOC monitoring by Brewer spectrophotometer in Southern Space Observatory SSO/CRS/INPE – MCTI (29.4 °S; 53.8°O; 488.7m) station for more than twenty years (1992 - 2014). Through this two-fold approach, we search for confronting our design goals with results from an empirical research proposed by the case study. Thus, we deploy the OSE detection algorithm over different scenarios representing edge and pervasive computing networks, both to validate the algorithm and to infer its execution performance.

This remain of this paper is organized as follows: we start presenting big data, the limitations of traditional computing platforms and the notions of edge computing and pervasive grids. The next section introduces the distributed computing platform CloudFIT and explain its main features. This section is followed by a case study that illustrates the usage of CloudFIT with a real application. Finally, we conclude this paper and explore future research directions.