Field-programmable gate array based fog analytic node architecture with reconfigurable application plane

Abstract

Fog computing extends computer networks by embedding analytics into intermediary network devices that are closer to data sources. Fog computing processes sensors data locally, conserves network bandwidth, improves process efficiency, and protects information privacy. However, a fog computing node demands high throughput, low latency, and high energy efficiency in data and packet processing. Besides, a fog node needs high architectural flexibility to enable timely functional updates for maintaining the relevancy of hosted analytics. This thesis proposes a fieldprogrammable gate array (FPGA) based fog node architecture with reconfigurable application plane for fog analytics. The proposed fog node’s application plane can be remotely reconfigured at run-time to enable dynamic redeployment of various fog analytics. The reconfigurable application plane allows run-time queuing scheme alteration to prioritize certain network ports and supports scaling on processing entity to cope with increased application loads.