The work described in this paper presents Graph-to-P4, a toolset for generating P4 boilerplate code from parse graphs. The purpose of such a toolset is to provide a higher level of abstraction to program pipelines in order to fill the knowledge gap that some students may encounter when diving into P4. While P4 is not difficult to learn, Graph-to-P4 offers the possibility first draw parse graphs, generate P4 starter code and finally let programmers develop the specific data plane behavior (actions, tables, etc.). We have experienced that this toolset offers a smooth introduction to P4 and the intrinsic networking concepts attached to it. We show that the toolset has potential to provide a larger set of high-level abstractions (tables, actions, etc.)and an extension to provide a two-way code transformation.

A standard priority-queuing system is capable of arranging packets with different traffic classes to guarantee a relatively low latency for the high priority traffic. However, in practical cases, severe delay may be caused by starting a large, low-priority frame ahead of a time-critical frame. In this paper, interspersed express traffic is evaluated, which enables preemption of non-time-critical transmission, in particular, the preemptive queuing system allows the cut-through transmission for critical traffic and minimizes the jitter. We analyse the performance of packet preemption through a system level simulation in Riverbed Modeler. The simulation is complemented by numerical analysis which provides the average queuing delay for both types of traffic (preemptable and express). Furthermore, the paper describes an approach to implement the packet preemption solution on an FPGA in VHDL, which illustrates the complexity of hardware implementation.