Publicação
Orchestration strategies for time-constrained smart city services
| Resumo: | As the scale and complexity of a smart city infrastructure increases, so does the need for comprehensive orchestration strategies that maximize automation, efficiency, fault-tolerance, flexibility, and observability. However, this type of infrastructure inherently presents some challenges that must be considered before opting to deploy one of the existing orchestration solutions. For some types of critical smart city services, maintaining an upper bound for the time required for producing the results can be more important than the results themselves. Without careful planning and tuning, these timing constraints can frequently be violated due to interference from other applications running on the resource-constrained nodes that typically comprise a smart cities’ edge computing infrastructure. Deploying one of the freely available orchestration solutions, without modification, can actually exacerbate these issues by introducing overhead and mismanaging deployments due to not being able to discriminate between critical and low priority services. This dissertation proposes a three-pronged approach to address these issues. Firstly, the applications themselves should be designed using best practices that increase orchestration efficiency and minimize processing and communication latency. As a case study, a new ETSI C-ITS protocol stack was implemented using a microservice architecture that represents a paradigm shift in the way that C-V2X applications are designed, in how they interact with each other and with the VANET, and also in how they can be orchestrated in an efficient manner. Secondly, critical applications should be executed using specialised configurations that mitigate interference by other processes, such as Real-Time scheduling. A comprehensive experimental evaluation was performed in order to study the effects of several different runtime configurations on the performance of an application in an edge computing single board computer. Finally, smart city orchestration tools should be extended with domain-specific logic, new functionalities and dynamic orchestration that promote a more optimal placement of services in worker nodes and the long-term stability of the cluster. In order to validate their implementation, each component was individually tested on a virtualised cluster using bespoke methodologies for each one. Results show that all the components performed within the levels that were expected and, in general, exhibited significant improvements relative to Kubernetes’s default behaviour. |
|---|---|
| Autores principais: | Rosmaninho, Rodrigo de Larmand Alvim Leal |
| Assunto: | Smart city Edge computing Vehicular ad-hoc networks Orchestration Kubernetes Real-Time scheduling |
| Ano: | 2023 |
| 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: | As the scale and complexity of a smart city infrastructure increases, so does the need for comprehensive orchestration strategies that maximize automation, efficiency, fault-tolerance, flexibility, and observability. However, this type of infrastructure inherently presents some challenges that must be considered before opting to deploy one of the existing orchestration solutions. For some types of critical smart city services, maintaining an upper bound for the time required for producing the results can be more important than the results themselves. Without careful planning and tuning, these timing constraints can frequently be violated due to interference from other applications running on the resource-constrained nodes that typically comprise a smart cities’ edge computing infrastructure. Deploying one of the freely available orchestration solutions, without modification, can actually exacerbate these issues by introducing overhead and mismanaging deployments due to not being able to discriminate between critical and low priority services. This dissertation proposes a three-pronged approach to address these issues. Firstly, the applications themselves should be designed using best practices that increase orchestration efficiency and minimize processing and communication latency. As a case study, a new ETSI C-ITS protocol stack was implemented using a microservice architecture that represents a paradigm shift in the way that C-V2X applications are designed, in how they interact with each other and with the VANET, and also in how they can be orchestrated in an efficient manner. Secondly, critical applications should be executed using specialised configurations that mitigate interference by other processes, such as Real-Time scheduling. A comprehensive experimental evaluation was performed in order to study the effects of several different runtime configurations on the performance of an application in an edge computing single board computer. Finally, smart city orchestration tools should be extended with domain-specific logic, new functionalities and dynamic orchestration that promote a more optimal placement of services in worker nodes and the long-term stability of the cluster. In order to validate their implementation, each component was individually tested on a virtualised cluster using bespoke methodologies for each one. Results show that all the components performed within the levels that were expected and, in general, exhibited significant improvements relative to Kubernetes’s default behaviour. |
|---|