ABSTRACT
In recent years, more and more devices are connected to the network, generating an overwhelming amount of data. This term that is booming today is known as the Internet of Things. In order to deal with these data close to the source, the term Edge Computing arises. The main objective is to address the limitations of cloud processing and satisfy the growing demand for applications and services that require low latency, greater efficiency and real-time response capabilities. Furthermore, it is essential to underscore the intrinsic connection between artificial intelligence and edge computing within the context of our study. This integral relationship not only addresses the challenges posed by data proliferation but also propels a transformative wave of innovation, shaping a new era of data processing capabilities at the network's edge. Edge devices can perform real-time data analysis and make autonomous decisions without relying on constant connectivity to the cloud. This article aims at analysing and comparing Edge Computing devices when artificial intelligence algorithms are deployed on them. To this end, a detailed experiment involving various edge devices, models and metrics is conducted. In addition, we will observe how artificial intelligence accelerators such as Tensor Processing Unit behave. This analysis seeks to respond to the choice of a device that best suits the necessary AI requirements. As a summary, in general terms, the Jetson Nano provides the best performance when only CPU is used. Nevertheless the utilisation of a TPU drastically enhances the results.
ABSTRACT
Development and operations (DevOps), artificial intelligence (AI), big data and edge-fog-cloud are disruptive technologies that may produce a radical transformation of the industry. Nevertheless, there are still major challenges to efficiently applying them in order to optimise productivity. Some of them are addressed in this article, concretely, with respect to the adequate management of information technology (IT) infrastructures for automated analysis processes in critical fields such as the mining industry. In this area, this paper presents a tool called Pangea aimed at automatically generating suitable execution environments for deploying analytic pipelines. These pipelines are decomposed into various steps to execute each one in the most suitable environment (edge, fog, cloud or on-premise) minimising latency and optimising the use of both hardware and software resources. Pangea is focused in three distinct objectives: (1) generating the required infrastructure if it does not previously exist; (2) provisioning it with the necessary requirements to run the pipelines (i.e., configuring each host operative system and software, install dependencies and download the code to execute); and (3) deploying the pipelines. In order to facilitate the use of the architecture, a representational state transfer application programming interface (REST API) is defined to interact with it. Therefore, in turn, a web client is proposed. Finally, it is worth noting that in addition to the production mode, a local development environment can be generated for testing and benchmarking purposes.
Subject(s)
Artificial Intelligence , Computational Biology , Big Data , Computers , Humans , SoftwareABSTRACT
In the smart city context, Big Data analytics plays an important role in processing the data collected through IoT devices. The analysis of the information gathered by sensors favors the generation of specific services and systems that not only improve the quality of life of the citizens, but also optimize the city resources. However, the difficulties of implementing this entire process in real scenarios are manifold, including the huge amount and heterogeneity of the devices, their geographical distribution, and the complexity of the necessary IT infrastructures. For this reason, the main contribution of this paper is the PADL description language, which has been specifically tailored to assist in the definition and operationalization phases of the machine learning life cycle. It provides annotations that serve as an abstraction layer from the underlying infrastructure and technologies, hence facilitating the work of data scientists and engineers. Due to its proficiency in the operationalization of distributed pipelines over edge, fog, and cloud layers, it is particularly useful in the complex and heterogeneous environments of smart cities. For this purpose, PADL contains functionalities for the specification of monitoring, notifications, and actuation capabilities. In addition, we provide tools that facilitate its adoption in production environments. Finally, we showcase the usefulness of the language by showing the definition of PADL-compliant analytical pipelines over two uses cases in a smart city context (flood control and waste management), demonstrating that its adoption is simple and beneficial for the definition of information and process flows in such environments.
