Publicação
Container orchestration at the edge: a comparative study of solutions for IoT workloads
| Resumo: | The growth of edge computing in smart city environments has increased the need for efficient orchestration technologies that can manage workloads close to where data is created. This dissertation, developed in collaboration between the University of Aveiro and Ubiwhere, focuses on finding the most suitable orchestration solution for edge deployments. The main goal is to give an informed recommendation on which technology best supports Ubiwhere’s workloads, with attention to performance and resource efficiency. Four Kubernetes-based orchestration solutions were evaluated: K8s, K3s, MicroK8s and KubeEdge. NVIDIA Deepstream was used as a representative workload for edge scenarios. For each orchestrator, a two-node cluster was created, and performance metrics such as CPU usage, RAM usage, GPU usage, deployment time and Deepstream’s FPS output were collected using Prometheus, Node Exporter and Grafana. The results show that K3s performed the best, especially in CPU, RAM and deployment time, while keeping similar GPU and FPS performance. These results indicate that K3s is the most effective option for edge workloads. |
|---|---|
| Autores principais: | Mourão, João dos Anjos da Silva |
| Assunto: | Edge computing Smart cities Orchestrators Internet of things |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Aveiro |
| Idioma: | inglês |
| Origem: | RIA - Repositório Institucional da Universidade de Aveiro |
| Resumo: | The growth of edge computing in smart city environments has increased the need for efficient orchestration technologies that can manage workloads close to where data is created. This dissertation, developed in collaboration between the University of Aveiro and Ubiwhere, focuses on finding the most suitable orchestration solution for edge deployments. The main goal is to give an informed recommendation on which technology best supports Ubiwhere’s workloads, with attention to performance and resource efficiency. Four Kubernetes-based orchestration solutions were evaluated: K8s, K3s, MicroK8s and KubeEdge. NVIDIA Deepstream was used as a representative workload for edge scenarios. For each orchestrator, a two-node cluster was created, and performance metrics such as CPU usage, RAM usage, GPU usage, deployment time and Deepstream’s FPS output were collected using Prometheus, Node Exporter and Grafana. The results show that K3s performed the best, especially in CPU, RAM and deployment time, while keeping similar GPU and FPS performance. These results indicate that K3s is the most effective option for edge workloads. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.