We introduce FAST (fine-grained, automatic state tracking and execution), to provide a new function store abstraction in OpenDaylight. Implemented as a shim layer on top of the native OpenDaylight MD-SAL datastore API, FAST provides automatic data-dependency tracking, automatic re-execution upon data changes, and re-execution scheduling. We will demonstrate the simplicity of programming using FAST (e.g., substantial simplification of implementing intents), present its design choices, and discuss potentials to extend FAST to provide additional benefits such as data provenance, through which one can trace the origin of mis-configurations, and shadow application diagnosis, through which one may test applications in a virtualized setting without affecting the real network. We will also discuss the limitations, extensions of FAST and seek collaboration from the community.
