The overwhelming amount of data that needs to be processed nowadays has driven the design of increasingly complex distributed systems. This complexity is further exacerbated by new decentralised approaches, which process the data near where it is generated, such as Fog or Edge computing.
Having full control of these infrastructures becomes a challenge even for the most experienced administrators, as there are many heterogeneous technologies and factors involved. Usually, administrators follow a process that involves using monitoring tools and browsing through logs in order to nd insights that explain events happening in the system. Instead, this cumbersome process can be partially or totally automatised through the use of arti cial intelligence techniques (AI) that extract these insights from all the incoming monitored information. In this thesis we propose a series of AI models that are able to solve some of the common problems that administrators nd in these kind of systems. Namely, we focus on providing observability for a Fog computing infrastructure, optimising the task parallelisation of Big Data jobs and performing root cause analysis for microservice architectures. Through a series of experiments, we demonstrate how the AI models can be used to increase the performance and reliability of data analytics jobs and the new technologies that power them.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados