Today the microservice architectural style is being adopted by many key technological players such as Netflix, Amazon, The Guardian. A microservice architecture is composed of a large set of small services, each running in its own process and communicating with lightweight mechanisms (often via REST APIs). If on one side having a large set of independently developed services helps in terms of developer productivity, scalability, maintainability, on the other side it is very difficult to have a clear understanding of the overall architecture of a microservice-based software system, specially when the deployment and operation of the involved microservices evolves at run-time. In this paper we present MicroART, an architecture recovery approach for microservice-based systems. By using Model-Driven Engineering techniques, we leverage a suitably defined domain-specific language for representing the key aspects of the architecture of a microservice-based system and provide a tool-chain for automatically extracting architecture models of the system. The only inputs of MicroART are: (i) a GitHub repository containing the source code of the system and (ii) a reference to the container engine managing it. We validated MicroART on a publicly available benchmark system, with promising results.

Microservice-based systems are characterised by a multitude of small services, each running in its own process and communicating with lightweight mechanisms. The microservice architectural style strongly encourages high decoupling among microservices in order to ease their independent deployment, operation, and maintenance. However, there are situations in which having a global overview of the system is fundamental. In this paper we present the first prototype of our Architecture Recovery Tool for microservice-based systems called MicroART. MicroART following Model-Driven Engineering principles, is able to generate models of the software architecture of a microservice-based system, that can be managed by software architects for multiple purposes.

Recently proposed model repositories aim to support specific needs--for example, collaborative modeling, the ability to use different modeling tools in software life-cycle management, tool interoperability, increased model reuse, and the integration of heterogeneous models.

Metamodels can be considered one of the cardinal concepts of Model-Driven Engineering, one which a number of coordinated entities such as models, transformations and tools, are dependent on. Analogously to any software artifact, metamodels are equally prone to evolution during their lifetime. As a consequence, whenever a metamodel changes, any related entity must be consistently adapted for preserving its wellformedness, consistency, or intrinsic correctness. This paper discusses the problem of co-adapting models, transformations, and tools. Different aspects are taken into account and a prospective and unifying characterization is given with the intent of clarifying the main difficulties and outline the basic requirements for possible solutions. In this respect, EMFMigrate a comprehensive approach to the metamodel co-evolution problem is proposed.

Manual classification methods of metamodel repositories require highly trained personnel and the results are usually influenced by the subjectivity of human perception. Therefore, automated metamodel classification is very desirable and stringent. In this work, Machine Learning techniques have been employed for metamodel automated classification. In particular, a tool implementing a feed-forward neural network is introduced to classify metamodels. An experimental evaluation over a dataset of 555 metamodels demonstrates that the technique permits to learn from manually classified data and effectively categorize incoming unlabeled data with a considerably high prediction rate: the best performance comprehends 95.40% as success rate, 0.945 as precision, 0.938 as recall, and 0.942 as F1 score.