Seleção dinâmica de microsserviços para coreografias

• Autor(a) principal: Silva, Douglas Bernardes
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFG
• Texto Completo: http://repositorio.bc.ufg.br/tede/handle/tede/10107
• topic: Coreografias de microsserviços; Coreografias adaptativas; Sistemas emergentes; Microsserviços; Microservices' choreography; Adaptive choreography; Emergent systems; Microservices; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

Dynamic and heterogeneous environments, such as smart cities, require increasingly flexible and interoperable software systems. Some requests, however, can not be met by simple collaboration between the available systems, leading to the need for new systems to be deployed at runtime. Approaches for building emergent systems have been proposed in order to use a set of small available software components to autonomously build this type of system. Microservice architecture characteristics make it an interesting candidate for this proposition. Microservices, however, must have a specific, well-defined purpose, and must be coordinated and combined to reach more complex results. Choreographies represent an interesting approach to do so. However, the selection of appropriate microservices to perform the roles specified in a choreography will only be effective if it ensures user QoS requirements, and that is not a trivial task since each candidate microservice's QoS tends to vary considerably by their reuse and inconstant execution context. It also means that design-time predictions tend to be ineffective. This dissertation proposes a QoSsensitive oriented selection of microservices during choreographies enactment in dynamic and heterogeneous environments. The proposal is to insert a new layer in the traditional architecture that provides support for choreographies, so that microservices selection ensures local QoS requirements for each functionality, at runtime, and enactment adapts "on the fly". A prototype of the proposed architecture was implemented for its evaluation, and the results indicate that the approach is effective and efficient when there is a large and multivariate set of microservices available, presenting superior performance in relation to the related works used for comparison.