Towards flexible in-network processing of data streams
Modern packet processing engines (PPEs) are faced with highly heterogeneous workloads driven by high volumes of end users and application types. A primary design challenge in this context consists in selecting and developing PPEs that scale application performance in a robust and cost-effective way providing the desired flexibility level and high performance. Interrelation among these objectives is non-trivial and definitely involves a certain conflict. In particular, flexibility is a driving objective to introduce new operational behaviors; from the other hand, performance and simplicity are constraining factors that ensure specific requirements. In this project we show that adding flexibility to the scheduling module and enhancing classification capabilities will allow not only consideration of user-defined objectives but also implementations of in-network processing of data streams.