Data centers are the invisible strings that hold things together for enterprises in nearly every vertical — from health care, finance, and manufacturing to energy, transportation, and logistics.

As enterprises put more focus on digital transformation and distributed cloud applications, innovation is proceeding at a rapid pace, and data center traffic is ramping up. Naturally, enterprises are looking for ways to increase the performance, scalability, and efficiency of their data center network operations.

However, in their quest for speed and agility, enterprises might be concerned they could be pushing their data centers to the breaking point, putting their critical operations at risk.

And, they are exactly right.

When it comes to data center performance for cloud, IoT, AI, 5G, or Industry 4.0 applications, good enough is no longer good enough. Enterprises need modern, open, and automated data center networking approaches that can support innovation and inspire confidence.

Changing requirements for enterprise data centers

Continuous network innovation helps enterprise networking teams address the needs of distributed cloud applications by boosting server and storage density, processing performance, and energy efficiency.

A data center software architecture that supports Kubernetes containers, microservices, and DevOps processes using continuous integration and continuous delivery (CI/CD) methods enables NetOps and cloud teams to develop and deliver applications quickly.

As an enterprise business grows, its data centers and cloud networks must support more devices, workloads, locations, capacity, and traffic. Many enterprises have modernized their data centers with leaf-spine architectures, also called data center fabrics. These architectures provide the scalability and bandwidth required within the data center and for interconnecting to their clouds.

But a key challenge for the NetOps and cloud teams is finding the time to develop the new applications and services they need to grow and transform the business while also performing routine daily operations, maintenance, and administrative tasks to keep the business running.

Suppliers of data center networking and operational tools have attempted to address this need by adding a range of automation capabilities to their platforms. They have also made their platforms more open and integrated with third-party applications — a much-needed evolution away from proprietary closed operating systems (OS).

But, while some leading platforms feature more open network operating systems (NOS) based on Linux, they don’t support easy customization, integration, or automation. It remains difficult for enterprises to adapt their data centers to new demands with speed and confidence in a partially open, automated environment.

A new approach to data center fabrics

The goal for enterprises is to gain competitive advantage by building intuitive, elastic data centers that can bring more scale and value and respond to new types of workloads with more agility.

Enterprises can’t achieve this goal with closed systems or cumbersome, partially open solutions. They need a generation of fully open data center networking solutions that make it easy for networking teams to design, deploy, adapt, operate, and automate data center network fabrics at scale.

But what does a next-generation data center fabric solution look like?

A next-generation solution starts with a truly open Linux-based NOS that provides tools for agile, flexible operations. The NOS must enable greater performance, scalability, and efficiency across all data center network operations.

An ideal NOS will provide data center network teams and application developers with:

• A cloud-native design that delivers programmability, flexibility, and resilient IP routing.
• An unmodified Linux kernel that can be used to build modular network applications that are isolated into their own failure domains
• A model-driven management architecture that ensures greater scalability and visibility and simpler operations
• A microservices-based, state-efficient design that supports hitless per-application upgrades, resilient networking, and high availability.
• A state-sharing, pub/sub architecture that uses protobufs and gRPC, and an impart database (IDB) that provides reliable, scalable, and secure communication channels.
• Easy interoperability through support for protocol stacks, such as Multiprotocol-Border Gateway Protocol (MP-BGP), Ethernet VPN (EVPN), and Virtual Extensible LAN (VXLAN).
• An open, scalable telemetry design interface for collecting streaming telemetry data without translation layers, which can be used to guide automated actions.
• Open programmable CLI with support for an open community ecosystem, which allows for fully customized operations.

Cloud-native tools designed for declarative, intent-based automation can help enterprises modernize and automate data center fabric design, deployment, and day-to-day operations. With a toolkit that takes advantage of open frameworks and microservices, enterprises can apply NetOps approaches to increase efficiency for all phases of data center fabric operations.

The introduction of new data center automations, applications, integrations, and workloads can, of course, create risk for enterprises. New data center fabric solutions would ideally provide a safe emulation environment, such as a data center network digital twin, where NetOps teams can test changes and see their effects before implementing them in the live network.

As such, enterprises will need high-performance hardware to get the most from a fully open NOS and automation toolkit. Data center switching platforms based on merchant silicon with a common hardware and software design and multiple form factors will give network teams maximum flexibility to address different data center applications.

These platforms should offer a range of Gigabit Ethernet interfaces including 400G to meet new speed and scalability demands, along with comprehensive Layer 2 switch and Layer 3 routing, quality of service, telemetry, security, and model-driven management capabilities.

How enterprises benefit from next-generation data center fabrics

A data center fabric built with a truly open NOS, flexible automation tools and high-performance hardware that’s scalable can help enterprises gain a competitive edge in the digital era.

It will make it easy for them to modernize their data center network operations to support any critical workload, accelerate application development, and maintain high application performance.

By taking advantage of these capabilities to build a high-performance, scalable, and efficient next-generation data center fabric, enterprises can create flexible design foundations for implementing and customizing network tools; simplify and automate every phase of data center fabric operations — from Day 0 design to Day 1 deployment and Day 2 operations and beyond — and quickly and safely develop, test, validate, and apply new configurations, automations, and services.

Houman Modarres is head of enterprise marketing at Nokia. He is passionate about the new ways networking, cloud, and analytics technologies can be applied to transform how enterprises run their businesses and optimize their outcomes. Previously, he held many positions in venture-backed startups as well as multinational firms. In every case, he considers himself very lucky to work with amazing engineers, customers, and mentors. Houman holds an MBA from UC Berkeley and a master’s in electrical engineering from Columbia University.