Decoupling algorithms and optimizations in network functions

Abstract

Network function virtualization promises a path to rapid innovation in networks. However, due to the complexity of developing these functions, innovations have been slow. Designing a network function is a daunting task that requires combining packet processing optimizations with the network function logic. It is not possible to ignore packet processing optimizations either: an optimized pipeline can have three times better performance than an unoptimized pipeline in our experiments. In this paper, we introduce NFMorph, a framework wherein the algorithms (i.e., the network function logic) are decoupled from the packet processing optimizations. Developers would specify the packet processing algorithms in a high-level language. The runtime then identifies the best set of optimizations on the algorithms. This is done based on the domain knowledge that operators provide as input to NFMorph as well as optimization templates we have developed for common NF primitives. NFMorph can also just-in-time reoptimize based on workloads and environment constraints.