Ushering in a new era of data center switching
The pandemic has focused attention on the importance of key infrastructure such as the public health care system, logistics and supply chains and our communications networks. All of them have been seriously tested and have so far held up remarkably well. Front-line health care workers have been the real heroes, but the internet, the cloud and supporting data centers have also played key roles in keeping us informed, entertained and productive throughout the pandemic.
The pandemic has accelerated what was already a significant trend: the crucial and expanding role that data center and cloud environments play in this digital age of 5G, IoT, AI and machine learning. Hyperscalers are the leading players in this trend, successfully building and operating data centers that support massive scale and efficient operations.
Adapting hyperscaler innovations for the mass market
Somewhat surprisingly, existing data center networking solutions do not reflect the experiences, practices and lessons learned by hyperscalers. Cloud builders depend heavily on edge cloud processing and cloud interconnect services, and need data center networks with improved scalability, flexibility and automation. And data center network operations teams, need to respond quickly and confidently to the demands of users while containing costs and improving efficiency. Meeting these needs is what motivated our approach to building Nokia’s new Data Center Switching Fabric solution.
Over the last two years, we worked with some of the industry’s largest cloud builders to evolve data center switching fabrics to bring hyperscaler innovations to the mass market — what we refer to as third-generation data center fabrics.
Existing data center switching solutions fall short
First-generation data centers were primarily designed to provide access to content and storage. Early first-generation data center switches evolved from LAN switches, so were fundamentally bridging and VLAN systems — a very limited foundation.
In the second generation, there was a shift to supporting cloud-native and microservices architecture, but there were limitations to controlling the underlying network. Second-generation data center switches are better but have differing support for the enhanced IP protocols used for leaf-spine fabric designs – such as Multiprotocol Border Gateway Protocol (MP-BGP), Ethernet VPN (EVPN) and Virtual Extensible LAN (VXLAN). Network operating system (NOS) architectures also vary widely and have limitations when it comes to scaling advanced protocols. Additionally, while many offer open interfaces, their development environments give limited direct access to the underlying network resources.
A new era of data center switching
As our customers kept asking for more direct access to network resources, we saw the need for a third-generation data center switching solution. Starting with a clean slate, we began at the network level and built up.
The result is a fully open, modern NOS and operations environment with all the capabilities of the system available for innovation. You can write your own applications and tools, or if you prefer, you can use ours, as the foundational networking functions are all there, robust and ready-to-go. All applications are peers and have full access to the system. What we have created, working closely with our customers, is a new approach that provides them with a data center fabric to support a future cloud-based world.
We identified three main areas that needed to be addressed:
- Keep pace with rapidly shifting, global-scale application demands, which means that the switching fabric needs to be simpler, more automated and able to respond rapidly.
- Take control of the network, which means the ability to develop network applications and customize and adapt tools as needed.
- Adopt automation to simplify data center fabric design, deployment and operations.
The Nokia Data Center Fabric solution is designed to meet these requirements. We have developed a new Network Operating System (NOS) called Service Router Linux (SR Linux) that is built on an open Linux® kernel. This robust, open and flexible platform implements a ground-up model-driven foundation and supports robust protocols including EVPN/VXLAN and MP-BGP, leveraging the resilient and field-proven IP routing protocols from our Service Router Operating System (SR OS).
SR Linux is supported on a portfolio of scalable data center switching platforms designed for high-performance leaf and spine and data center interconnect roles.
Data center NetOps revisited
To complete the picture, we implemented a data center network operations (NetOps) system in response to customer requests to help automate their Day 0, Day 1 and Day 2+ fabric operations. Many customers lack this capability today or it is complex and cumbersome to implement with their existing data center switching solutions.
Our design philosophy was to develop a complete set of capabilities to support all NetOps goals. At the NOS level, SR Linux is natively model-driven and is immediately ready for streaming telemetry without requiring any translation layers. Operators can completely customize the way the CLI operates to allow advanced commands and logic. Our NetOps Development Kit (NDK) enables application developers to leverage SR Linux’s underlying model-driven architecture, with a simple, clean, decoupled integration.
Our solution includes a fabric automation system, which complements SR Linux by providing tools for operations teams to easily design, deploy, test and operate data center switching fabrics at speed and scale with full control. The system supports declarative, intent-based design approaches, and provides a digital sandbox, a virtual twin that mirrors the actual fabric. This allows operations teams to test applications and reduce risk before going to live deployment. Extensive telemetry and real-time monitoring alert teams whenever fabric operations deviate from operational intent.
We believe our SR Linux NOS and fabric automation system enables a unique approach to enhancing data center NetOps, and we’re excited about the new solution. And by deploying and testing it with some of the largest cloud providers, we have seen that they are too. Cloud builders of all sizes now have a truly open and powerful NOS for data center and cloud environments that enables them to manage, not only what the pandemic has thrown at them, but also exciting future services, from 5G to Industry 4.0.
Watch out for upcoming blogs about our next-generation solution covering the SR Linux NOS, data center switching platforms and the fabric automation system.
For insights into the design, architecture and additional information on our Data Center Switching Fabric solution, please watch our Networking Field Day 25 (NFD25) - Data Center Switching overview presentation and listen to this Packet Pushers podcast - and check out our Data Center Fabric solution website.
The registered trademark Linux® is used pursuant to a sublicense from the Linux Foundation, the exclusive licensee of Linus Torvalds, owner of the mark on a worldwide basis.