Learning to Adapt – Software Engineering for Uncertainty
Deepali Kholkar, Suman Roychoudhury, Vinay Kulkarni, Sreedhar Reddy · 2022
Modern businesses are being subjected to an unprecedented variety of change drivers that cannot be predicted such as new regulations, emerging business models, and changing needs of stakeholders. This creates new demands on enterprises to meet stated goals in a dynamic and uncertain environment that translate to demands on the enterprise’s software systems. Software systems however are currently designed to deliver a fixed set of goals and assumed to operate in a static environment, falling short in addressing the need for continuous adaptation under uncertainty. State-of-the-art adaptation architectures like MAPE-K have been applied to meeting non-functional requirements in a dynamic environment using a static repository of knowledge. This paper articulates the need for architecting software systems that learn from their own operation to dynamically extend existing knowledge, and utilize the knowledge to meet stated functional goals in an uncertain environment.