Improving container deployment latency in distributed edge infrastructures
New services, such as augmented reality and natural language processing, require some network and processing thresholds that aren’t possible with Cloud Computing. New paradigms near the end-user, like Multi-Access Edge Computing and Fog Computing, or generically speaking Edge Computing, emerged to b...
| Autor: | |
|---|---|
| Formato: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2021 |
| País: | Brasil |
| Recursos: | Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS) |
| Repositorio: | Biblioteca Digital de Teses e Dissertações da PUC_RS |
| Idioma: | inglés |
| OAI Identifier: | oai:tede2.pucrs.br:tede/10148 |
| Acesso em linha: | https://tede2.pucrs.br/tede2/handle/tede/10148 |
| Access Level: | acceso abierto |
| Palavra-chave: | Container Management Edge Computing Orchestration Container Scheduling Gerenciamento de Contêineres Computação na Borda Orquestração Escalonamento de Contêineres CIENCIA DA COMPUTACAO::TEORIA DA COMPUTACAO |
| Resumo: | New services, such as augmented reality and natural language processing, require some network and processing thresholds that aren’t possible with Cloud Computing. New paradigms near the end-user, like Multi-Access Edge Computing and Fog Computing, or generically speaking Edge Computing, emerged to bring these requisites to such applications. However, this new paradigm presents several challenges, such as the fast and continuous provision of applications on geographically distributed heterogeneous devices at the edge, often with constraint resources. Currently, there are few strategies to decrease application deployment time in container-based infrastructure. However, the specificities of an Edge scenario and the several components presents in these topologies have several points that need to be optimized before a large adoption of this paradigm. With that in mind, this thesis presents four main contributions. First, the development of an event-driven simulator to edge container orchestration. After, we give three contributions on distinct components, a fluid communities placement for the container registries, a new priority to the kube-scheduler based on the network availability, and a new Deployment SLA-driven scheduler using a multiobjective genetic algorithm. |
|---|