Adapting embeded Tomcat to develop event-driven serverless functions





Publicado en

Actas de las XVII Jornadas de Ciencia e Ingeniería de Servicios (JCIS 2022)

Licencia Creative Commons


The deployment of software has suffered major transformations in the last decade. The path has led us to the advent of serverless, where resources are not required to be allocated in advance, scale to zero if there are no requests thus paying only for the time the functions are running. If functions are allowed to scale to zero, it is possible that the first call will have to wait for the activation of the function and then for the generation of the response, this is known as cold start. Another important consideration, related to the previous one, is that the orchestrator will schedule the execution in a node, and if that node does not have the container image it has to be downloaded before the contanier can be started. Therefore, startup time and image size matters. This work describes an adaptation of embeded Tomcat to ease the development of event-driven functions. This approach can lead to small images, compared with other frameworks, and thus smaller response times in cold starts. Besides, smaller images and applications with less memory consumption are suitable to run in edge devices with smaller computational capabilities and memory constraints. To test the proposed modification, a function that uses YOLOv3 to detect objects in images, has been developed and deployed on Knative.


Acerca de Gutiérrez-Aguado, Juan

Palabras clave

DNN, Events, Function As A Service, Knative, Serverless
Página completa del ítem
Notificar un error en este artículo
Mostrar cita
Mostrar cita en BibTeX
Descargar cita en BibTeX