LightKone Reference Architecture (LiRA)

The LightKone Reference Architecture (LiRA) presents a novel edge reference architecture that takes advantage of decentralized lateral data sharing and convergent vertical data semantics across a myriad of different edge resources. These two principles are key to bridge the existing gaps in current edge-based proposals, namely when considering fog reference architectures, and in particular, by removing the need of centralized lateral data sharing among components that exist in close vicinity and promoting unified data sharing semantics across components in the edge and the core of the system, respectively. LiRA achieves this by moving its focus to application level semantics and exploiting them to enable convergence properties for decentralized data management mechanisms. LiRA is supported by proved and sound techniques like Conflict-free Replicated Data Types (CRDTs) and Transaction Causal Consistency (TCC). These are key to achieve fundamental properties in edge-based solutions such as: autonomy, availability, consistency, and robustness. LiRA is compatible with, and complementary to, cutting edge/fog standards such as the OpenFog Reference Architecture [5, 12]. To show the practical feasibility of LiRA, this paper presents a reference implementation of LIRA, called i-LiRA, composed of a coherent collection of software artifacts and components, that ease the use of the concepts underlying to LiRA in the development of practical applications that take advantage of edge computing. The paper concludes with a discussion on how the artifacts of i-LiRA were leveraged to implement four different edge computing case studies across different application domains.