Flexilient End-to-End Architectures

Highly intelligent, massively connected, autonomous systems featuring state-of-the-art technologies offer as many opportunities as challenges. The Internet-of-Things paradigm enters all areas of life. However, it is not enough to just provide intelligence or autonomy to systems. They must be able to connect to other systems, provide services and end-to-end communications, adapt to changing needs and be dependable at the same time. The challenge is resilience - the persistence of dependability when facing changes - and how it can be defined in the context of end-to-end architectures, which consist of many layers of components, both software and hardware, which in turn can have different safety, availability and dependability requirements. We propose the term flexilience - the combination of flexibility, intelligence and resilience to describe this area of conflict. Flexibility thrives to constantly and perfectly adapt to the present conditions, while intelligence continuously increases the systems' cognitive capabilities and resilience ensures its dependability in changing conditions. Therefore, flexilience is persistent dependability and optimized performance in cognitive systems when facing changes. In intelligent autonomous systems, considering only worst-case scenarios during the design phase would result in dramatically limited performance. For example, potential cloud or edge services in such scenarios could not be used for any safety-related functions. The actual situation and risk should thus be taken into account. In this paper, we present a novel approach for designing and managing such systems at runtime that allows safety aspects to be evaluated and guaranteed not only during the design phase and for worst-case scenarios, but also at runtime in line with the current situation. As a result, we can move functions, including those that are safety related, to the cloud or edge for improved performance.