Demand for network bandwidth is accelerating at an extraordinary pace, challenging network operators of all kinds to keep up with customer needs. Implementing network functions in software, known as network function virtualization (NFV), is one way to help keep pace and is growing in use. To ensure NFV can meet the performance requirements of next-generation applications, Intel is assembling an ecosystem of partners to offer an approach based on Intel field programmable gate arrays (FPGAs).
Global IP traffic is expected to reach nearly 50,000PB per month by 2021, up from less than 12,000 PB per month in 2017, according to the Cisco Visual Networking Index. Global IP traffic will increase nearly threefold from 2016 to 2021, and 127-fold from 2005 to 2021, Cisco predicts.
In addition to traffic, the number of subscribers is increasing, along with quality of service (QoS) requirements—meaning customers expect ever faster networks with low latency. Those who own and operate networks, or make devices that deal with network traffic, are naturally interested in new approaches to deal with explosive data growth and stringent customer demands.
Any solution must be able to deliver data security and integrity, while ensuring confidentiality and authentication of all data communications. That requires tools such as IPSec, a key ingredient for data protection but one that consumes significant CPU resources when performing data encryption and decryption.
NFV can help address such issues by supporting functions including virtual Broadband Network Gateways (vBNGs) and IPSec. To accelerate NFV adoption, Intel is assembling its partner ecosystem—comprising OEMs, virtual network function (VNF) partners, ISVs, and systems integrators—and coming up with reference designs based on Intel FPGAs.
The idea is to use FPGAs to accelerate various workloads, providing a number of benefits, including:
Scalability and capability: FPGAs can be programmed to deliver a range of high-bandwidth, low-latency functions. At the recent Mobile World Congress event in Shanghai, Intel and Lenovo demonstrated how an FPGA-based Intel Programmable Acceleration Card accelerated vBNG QoS performance, delivering a superior video playback experience with no jitter as compared to a vBNG solution without hardware-based acceleration.
Programmability: Networks need to adapt to many new applications, rather than expect the applications to adapt to the networks. Because they are programmable, FPGAs are like a malleable glue that enables the same network to support an array of applications, delivering to each the performance it requires.
Diversity: By the same token, that programmability makes FPGAs ideally suited to be reprogrammed as necessary to address new requirements as they emerge.
With its network of partners, Intel is accelerating the creation of an end-to-end NFV ecosystem that will deliver on the promise of FPGAs in addressing the explosion of network traffic.
Click here to learn more about how FPGAs can help.