Evelyne Roch

Uniform handling and abstraction of NFV hardware accelerators

This article outlines some NFV topics related to implementing VNF over NFVI where HWA may be needed. The document highlights the commonalities and differences among various HWA approaches and suggests uniform handling, common architecture, and abstraction layer for the hardware acceleration components. This uniform handling will allow deployment of various hardware accelerators within NFVI and facilitate interoperability between various VNFs and HWAs.

Accelerating NFV with fast path offloading

In this paper, we propose Fast Path Offloading (FPO), a new framework that accelerates the performance of many virtual Network Functions. FPO enables the Network Functions to offload the majority of the traffic processing to the underlying network. We implemented FPO and used it to accelerate the performance of HA-proxy, an open source L7 load balancer.