Keywords
Microservices, Service Discovery, LSTM Networks, Reinforcement learning, Q-learning, self-adaptation, multi-level adaptation, IoT
Objective
The objective of this focus area is two fold: i) leverage combination of machine learning techniques to optimize the service discovery of microservices by considering QoS of the microservice instances such as response time, utilization, etc. and context parameters of the instances such as energy consumption, location, time, etc; ii) make use of the expressiveness provided by formal goal models along with the capabilities offered by deep learning techniques such as LSTM to develop a multi-level adaptation approach for optimizing QoS of microservices and IoT devices by ensuring that the Quality of Experience (QoE) of users is guaranteed in Microservice-based IoT systems.
Description
Microservice architectures (MSA) have become enormously popular since traditional monolithic architectures no longer meet the needs of scalability and the rapid develop- ment cycle of modern software systems. The success of large companies (Netflix among them) in building and deploying services is also a strong motivation for other companies to consider making the change. The loosely coupled property of microservices allows the independence between each service, thus enabling the rapid, frequent, and reliable delivery of large, complex applications. This is evident from the fact that microservice-based architecture (MSA) is considered as one of the best possible solutions for architecting data-driven and event-driven systems like IoT.
However, like most transformational trends, architecting and implementing a microservice-based system poses its own challenges: Hundreds of microservices may be composed to form a complex architecture; thousands of instances of the same microser- vice can run on different servers; the number or locations of running instances could change very frequently. In addition, the set of service instances changes dynamically because of autoscaling, failures, and upgrades. Therefore, one of the challenges in a microservice architecture concerns how services discover, connect, and interact with each other. Consequently, elaborated service discovery mechanisms are required that not only takes into consideration the QoS of the microservice instances but also considers the contextual dimension which involves the location of the instances, energy consumed by the instances, time of invocation, etc.
Moreover, additional challenges arise when microservices-based solutions are applied to IoT systems as the devices themselves are subjected to different uncertainties. These refer to the evaluation and maintenance of the Quality-of-Service (QoS) characteristics of systems (e.g., performance and reliability) due to the uncertainties faced by IoT devices because of resource constraints (e.g., battery level, network traffic)