FPGA based Virtual Fabric for Distributed FOG/Edge Computing

5G and IoT infrastructure and new applications are expected to generate massive amounts of data, thereby driving changes to network architecture. Data processing and content distribution will be forced closer to the network edge, driving the need for much more efficient and programmable edge compute infrastructure. Traditionally, such edge compute infrastructure needs have been met with Field Programmable Gate Arrays (FPGAs) providing a reprogrammable protocol handling layer (at PHYs or NICs), and generic CPUs (possibly augmented with GPUs) interconnected through a largely static network switching fabric. The objective of this project is to investigate approaches and develop a software tool kit for managing and programming a fabric of interconnected FPGAs. The FPGA fabric (i.e. multiple FPGAs interconnected to appear as a single large virtual FPGA) is intended to absorb functions that are currently spread across multiple elements: protocol adapters, switching ASICs, networking functions, data processing CPU/GPU functions, etc. The project is expected todeliver a set of algorithms, techniques and tools for adaptive and optimized sizing, placement and chaining of required functions inside the FPGA fabric.