High Performance Computing and Network Function Virtualization: A major challenge towards network programmability

Abstract

The combination of emerging concepts, such as Fog Computing, Software Defined Networking and Network Function Virtualization will contribute to transform the network into a flexible and dynamical continuum of resources, on which services can be easily deployed and managed (network programmability). In this scenario, a major topic is the use of ad hoc h/w accelerators, in particular to provide the network with general purpose High Performance Computing capabilities, whose application has been so far limited to few highly specialized contexts, such as graphics or scientific computing. However, the use of h/w acceleration is generally considered in contrast with the main NFV concepts, and a major obstacle towards network programmability. This paper argues that, on the contrary, h/w acceleration and virtualization technologies are fully compatible, and together can greatly enhance the network infrastructure. In particular, the use of Graphic Computing Units in Virtual Network Functions is analyzed, and their flexible management is considered, within the innovative NFV architecture designed by the EU FP7 T-Nova project.