- author = "Fredy Humberto {Vera Rivera}",
- title = "Modelo Inteligente de Especificacion de la Granularidad de Aplicaciones Basadas en Microservicios",
- school = "Escuela de Ingenieria de Sistemas y Computacion, Facultad de Ingenieria, Universidad del Valle",
- year = "2021",
- address = "Cali, Colombia",
- keywords = "genetic algorithms, genetic programming, SBSE, Service-oriented systems engineering, Service computing, Software design, Software architecture, Micro-services granularity, Microservices decompositions, Software metrics, Microservice Backlog, Cargo Tracking, JPet Store, Sinplafut",
-
URL = "
https://bibliotecadigital.univalle.edu.co/bitstream/10893/19567/1/9702-V473.pdf",
- size = "152 pages",
-
abstract = "Microservices are an architectural and organizational
approach to software development in which applications
are composed of small independent services that
communicate through a well-defined application
programming interface (API). Many companies use
microservices to structure their systems, and
microservices architecture has also been used in other
areas such as the Internet of Things (IoT), edge
computing, cloud computing, autonomous vehicle
development, telecommunications, E-Health systems,
E-Learning, among others. A great challenge when
designing this type of applications is to find an
appropriate partition or granularity of the
microservices, a process that to date is performed and
designed in an intuitive way, according to the
experience of the architect or the development team.
The definition of the granularity of microservices is
an open research topic. There are no standardized
patterns, methods or models that allow defining how
small a microservice should be. The most used
strategies to estimate the granularity of microservices
are machine learning, semantic similarity, genetic
programming, and domain engineering. In this doctoral
research work an intelligent model was proposed to
specify and evaluate the granularity of the
microservices that are part of an application,
considering some characteristics such as cognitive
complexity, development time, coupling, cohesion, and
its communication.
In chapter one the theoretical foundations are presented, the research problem is solved: How to model and define the adequate size of a microservice, considering its properties, dependencies, cognitive complexity, coupling and cohesion? together with the research questions that help to solve it, the objectives and research methodology are also presented. Based on Design Science Research, a new practice was proposed, an intelligent model of specification of the granularity of the microservices called Microsevices Backlog, the phases and research methods that allowed to solve the research questions posed are also presented. Chapter two presents the state of the art and the related works to the present doctoral research thesis; the metrics that have been used to define and evaluate the granularity of microservices are also identified. In Chapter three, the process of developing of microservice-based applications is characterized, explaining its use in a case study called Sinplafut. Chapter four describes the Microservice Backlog, and presents the definition of each of its components, among which are: the parameterising component, the grouping component (a genetic algorithm and a semantic grouping algorithm based on unsupervised machine learning), the metrics evaluation component and the decomposition and candidate microservices comparison component. It also presents the mathematical formulation of the granularity of applications based on microservices. The fifth chapter presents the evaluation of the proposed practice. It was done in an iterative way using four case studies, two examples raised in the state of the art (Cargo Tracking and JPet-Store) and two real projects (Foristom Conferences and Sinplafut), the Microservices Backlog was used to obtain and evaluate the candidate microservices of the four applications. A comparative analysis was made against methods proposed in the state of the art and with the domain-driven design (DDD), which is the most used method to define the microservices of an application. The Microservices Backlog obtained a low coupling, high cohesion, low complexity and reduces the communication among the microservices, this compared with the proposals of the state of the art and with DDD. Finally, in chapter six we present the conclusions, contributions, limitations, and products obtained as result of this thesis",
-
notes = "is this GP?
In Spanish
Supervisor: CARLOS MAURICIO GAONA CUEVAS",
