Towards Performance-Aware Runtime Management of P4-based Heterogeneous Cloud Environments

The recent trend to deploy programmable packet processors in cloud environments enhances the packet processing capability without losing the flexibility to adapt the functions at runtime. In particular, distributed edge clouds can have a heterogeneous programmable processing substrate made up of different class of devices: CPUs, NPUs, FPGAs, etc. However, managing the embedding of processing workload into such heterogeneous programmable processing substrate, in particular deciding where to instantiate a certain function, is a non-trivial task with many decisive factors. In this paper, we propose a mathematical model for optimizing the embedding of Service Function Chains (SFCs) implemented in P4, while considering the functional and QoS requirements associated with embedding requests on one hand, and the various types of processing devices that have different properties in terms of processing delay and supported features on the other hand. To support satisfying delay QoS requirements, the problem formulation utilizes predeveloped performance models to predict the forwarding latency associated with different candidate embedding options. Furthermore, a greedy solution is proposed to solve the problem faster. Finally, a detailed numerical evaluation is conducted to study the impact of different system's characteristics and to evaluate the effectiveness of the proposed greedy solution.