knowledge centre

Acceleration technologies: realizing the potential of network virtualization

Gorkem Yigit Senior Analyst, Research
Caroline Chappell Research Director

"Acceleration technologies will be pivotal to network transformation by enabling operators meet the performance, latency, QoS, subscriber density and security requirements of existing and future applications with optimum TCO."

FREE WHITE PAPER

Acceleration technologies: realizing the potential of network virtualization

Operators are using industry-standard, commercial off-the-shelf (COTS) servers with general purpose CPUs to build their network function virtualization infrastructures (NFVIs) as they transition from closed, custom appliances to software-based network functions running on horizontal, open and highly automated platforms. The goal of network function virtualization (NFV) is still to have a homogenous, hyperscale, COTS architecture for all virtual network functions (VNFs). In practice, acceleration technologies (both fixed-function and programmable), in conjunction with tightly coupled general compute resources, can improve the overall networking and application performance of current VNFs to meet the quality of service (QoS), automation and security requirements of existing and future services, and improve capital and operational expenditures.

This whitepaper defines the various acceleration technologies that are typically explored, in conjunction with general-purpose hardware and software technologies, to maximise the performance and efficiency of NFV/SDN, 5G and artificial intelligence (AI)/machine learning (ML) applications. No single architecture can respond to or satisfy all use cases, but this paper seeks to outline some of the currently deployed acceleration technologies as well as those that are emerging, alongside performance enhancements achieved by general purpose CPUs. It presents performance results for accelerating virtualized security functions as a proof point for emerging acceleration technologies. Finally, it discusses the need for an acceleration technology abstraction layer and a uniform management framework to simplify and automate the operations of heterogenous acceleration resources with an ‘as-a-service’ model. It highlights industry initiatives such as OpenStack Cyborg and OPNFV DPACC, which aim to help operators to achieve these goals.