As enterprises readily consume cloud-based services, cloud providers have needed to build out the infrastructure that supports that consumption.
Most cloud providers have tackled this restructuring by scaling out their infrastructure, adding more servers to increase compute. But they also wanted to scale the services they offered, said Silvano Gai, author of Building a Future-Proof Cloud Infrastructure from Pearson. As they added more compute, they assessed how they could innovate what Gai called "the border between the server and the network" -- namely, the network interface card (NIC).
NICs have evolved over the years, gradually adding capabilities like packet filtering and traffic shaping. But the most recent innovation with NICs is the addition of processing and programmability to create what is called a smartNIC.
SmartNICs still perform the traditional NIC task of sending and receiving packets between networks and servers. But they can also take on other functions, such as processing, encryption and load balancing. Ideally, smartNICs can reduce server loads and, ultimately, improve performance. Many smartNICs also support the P4 programming language, which enables more specific data plane programmability for packet processing.
While most cloud providers have implemented smartNICs in their infrastructures, enterprises have been slower to adopt, largely because their networks don't require the same level of scalability. But enterprise smartNIC use cases are emerging, such as storage offloading, telemetry and private cloud support.
Below is an excerpt from Chapter 8, "NIC Evolution," from Building a Future-Proof Cloud Infrastructure. In this chapter, Gai discusses the role of smartNIC architecture in supporting distributed cloud infrastructure.
Explore Building a Future-Proof Cloud Infrastructure
Click here to read Chapter 8, "NIC Evolution."