RT info:eu-repo/semantics/masterThesis
T1 Computación de Altas Prestaciones en entornos contenerizados
A1 Pérez Znakar, Omar Patricio
A2 Máster en Ingeniería Informática
K1 Kubernetes
K1 Docker Hub
K1 Quay
K1 Machine Learning
K1 Python
K1 C
K1 MPI
K1 OpenMP
K1 Filtro Gaussiano
K1 Computación de Altas Prestaciones
AB Today Kubernetes deployments have had a major impact on web applicationsbased on microservices architectures. Its most notable potential is the ability toscale and manage the services associated with these applications that run inthe cloud, usually with container based technology. A wide variety of tools andenvironments are available that enable enhancements to Kubernetes deployments. However, one of the drawbacks that arise in this area is that it is notprecisely known which software has the ideal characteristics to provide betterperformance in specific problems or, if not, the execution is better on a systemwithout containerization. The objective of this project is to develop a comparison of runtime environments with and without containers in order to analyzewhether Kubernetes is a viable alternative to running on a server without containerization for High Performance Computing (HPC) workloads.In order to carry out this objective, the “GraphCloud ” application has beencreated for this project, which consists of a computer service that allows applications to be launched remotely from authorized clients. This application is underthe development stack called ” MEVN ” (Mongodb, Express, Vue.js and node.js)allowing, in addition, to apply various technologies in the field of High Performance Computing. At the same time, we must emphasize that an exhaustiveperformance analysis will be carried out on Machine Learning algorithms andGaussian filter on images, executing each one both sequentially and in parallel.The project has been developed using technology in three systems, in a personal computer (processor with 4 cores), in the IaaS of the University of La Laguna (corresponding to a structure of virtualized machines of a core) and, later,in the private server (Verode) of the research group ”High Performance Computing”(server with 40 integrated cores). The performance analysis we have conducted focuses on the latter system.
YR 2021
FD 2021
LK http://riull.ull.es/xmlui/handle/915/22678
UL http://riull.ull.es/xmlui/handle/915/22678
LA es
DS Repositorio institucional de la Universidad de La Laguna
RD 21-nov-2024