Self-balancing architectures based on liquid functions across computing continuums

Abstract

Scalable application development is highly influenced by two major trends - serverless computing and continuum computing. These trends have had little intersection, as most application architectures, even when following a microservices or function-based approach, are built around rather monolithic Function-as-a-Service engines that do not span continuums. Functions are thus separated code-wise but not infrastructure-wise, as they continue to run on the same single platform they have been deployed to. Moreover, developing and deploying distributed applications remains non-trivial and is a hurdle for enhancing the capabilities of mobile and sensing domains. To overcome this limitation, the concept of self-balancing architectures is introduced in which liquid functions traverse cloud and edge/fog platforms in a continuum as needed, represented by the abstract notion of pressure relief valves based on resource capacities, function execution durations and optimisation preferences. With CoRFu, a reference implementation of a continuum-wide distributed Function-as-a-Service engine is introduced and combined with a dynamic function offloading framework. The implementation is validated with a sensor data inference and regression application.