No abstract

Today many service-based systems follow the microservice architecture style. As microservices are used to build distributed systems and promote architecture properties such as independent service development, polyglot technology stacks including polyglot persistence, and loosely coupled dependencies, architecting data management is crucial in most microservice architectures. Many patterns and practices for microservice data management architectures have been proposed, but are today mainly informally discussed in the so-called "grey literature": practitioner blogs, experience reports, and system documentations. As a result, the architectural knowledge is scattered across many knowledge sources that are usually based on personal experiences, inconsistent, and, when studied on their own, incomplete. In this paper we report on a qualitative, in-depth study of 35 practitioner descriptions of best practices and patterns on microservice data management architectures. Following a model-based qualitative research method, we derived a formal architecture decision model containing 325 elements and relations. Comparing the completeness of our model with an existing pattern catalog, we conclude that our architectural decision model substantially reduces the effort needed to sufficiently understand microservice data management decisions, as well as the uncertainty in the design process.

Edge computing describes a paradigm for combining computational resources at the edge of the network with the cloud. Even though complementing the cloud with these resources provides benefits, e.g., low latency, it also introduces new challenges to the operational staff. Such challenges can be: deciding if the applications should be placed in the cloud or at the edge, and monitoring them at runtime to ensure that all the application requirements are met. This becomes more challenging when using microservices due to the complexity of the resulting placement problem. To mitigate such concerns, we introduce an automatic deployment framework along with a prototype implementation, called D-DAD. This framework provides a transparent (to the operational staff) way to deploy applications with respect to all their requirements-including the non-functional-using mechanisms for monitoring and adapting the deployments to the available resources in a cloud-edge environment. For evaluating our framework, we provide results from a series of experiments which show how the adaptation mechanism meets the application requirements, including a ∼90% reduction of CPU utilization violations, compared to using only the local resources.

One of the chief problems in software architecture is avoiding architecture model drift and erosion in all kinds of complex software systems. Microservice-based systems introduce new challenges in this context, as they often use a large variety of technologies in their latest iteration, and are changed and released very frequently. Existing solutions that can be used to reconstruct architecture models fall short in addressing these new challenges, as they cannot easily cope with continuous evolution, their accuracy is too low, and highly polyglot settings are not supported well. In this work, we report on a research study aiming to design a highly accurate architecture model abstraction approach for comprehending component architecture models of highly polyglot systems that can cope with continuous evolution. After analyzing the results of related studies, we found two possible architecture model abstraction approaches that meet the requirements of our study: an opportunistic, and a reusable semi-automatic detector-based approach. We have conducted an empirical case study for validation and comparison of the two approaches. We conclude that both detector approaches are feasible. In our case study, the reusable approach breaks even in terms of time and effort needed for establishing reuse, if modest reuse of detectors is possible, and is producing slightly more high quality and evolution-stable solutions than the opportunistic approach.

Many contemporary service-based systems follow the microservice approach, particularly in DevOps or continuous delivery contexts. They share a set of important tenets such as independent development and deployment, high releasability, polyglot technology support, and loose coupling. A number of best practices for microservice architectures have been codified as patterns, which embody those tenets. However, no real-world microservices system can support all patterns and practices well, but rather architectural decisions making trade-offs among them are needed. Conformance to the patterns and practices selected in such decisions is hard to ensure and assess automatically, especially in largescale, complex, and evolving systems. In this work, we propose a model-based approach based on generic, technology-independent metrics, tied to typical architectural design decisions in the microservice domain. With this approach we can measure conformance to the patterns and related tenets. We demonstrate and assess the validity and appropriateness of these metrics in performing an assessment of a system's conformance to patterns through statistical methods.