Data Stream Processing in Software Defined Networks: Perspectives and Challenges

摘要

The new paradigm of network softwarization is pushing programmability and programming abstractions as key elements at different levels on both data and control planes of the network. However, the availability of programmable abstractions and devices per se is not sufficient to guarantee high-speed processing rates to network applications without the adoption of efficient programming models and accelerating methodologies. The paper discusses the possible sources of computation bottlenecks and proposes Data Stream Processing (DaSP) as a viable programming model for a unified scheme of accelerating data elaboration over both fast and slow data paths. Perspectives and implications of the adoption of DaSP are presented along with possible research directions.