Life as a modern IT leader is filled with challenges, shifting priorities, and endless compromises. The North Star, in most cases, is to be an enabler of, not an impediment to, digital transformation. Of course, that is much easier said than done. At the same time, getting it right has never been more important especially when it comes to foundational infrastructure like the data center. In fact, according to IDC, Worldwide Datacenter 2019 Predictions, “A major threat to successful transformation for most businesses remains the failure of their IT organizations to convert from being the back-office enabler of internal business processes to playing a leading role as the engine powering digital business flows between people, things, and data.”

As many workloads move to the public cloud, others remain on premise at primary and backup data center locations due to concerns for cost, control, refactoring complexity, and data sovereignty. Many enterprises (and regional cloud/colocation providers) have at least two data center locations and with the rise of edge compute mini-data centers, more are likely to sprout up in locations such as the factory floor, oil rigs, central offices, or 5G base station colocation facilities.

This means that IT leaders, especially those in charge of the network, face an increasing number of potential challenges. While they want to be an enabler of digital transformation, their infrastructure is becoming more distributed and complex, increasing the chance of outages and making provisioning, management, and troubleshooting more time-consuming.

The promise of data center interconnect fabric

From a network perspective, data center interconnect (DCI) technology is used today to connect two or more geographically distributed data center locations. Traditionally, routers sit at the outside edge of the data center (DC Gateways) and use Layer 3 Internet Protocol (IP) connectivity over point-to-point dense wavelength division multiplexing (DWDM) transponders to route and transport traff­ic between geographically distributed data centers. This basic transport is required, but as edge compute starts to emerge and data centers become more distributed, it is not sufficient.

What is needed going forward is a data center interconnection fabric (DCIF) that can stretch across multiple sites, not just to the edge of the data center, but actually reach inside the data center down to every top of rack switch (leaf). The fabric should have the following attributes:

• The fabric makes N switches look like one logical switch through software-defined network (SDN) control
• Control of the entire fabric from any switch via CLI or through a standard API like REST

Configuration changes are populated to all switches to simplify operations, homogenize configurations, and reduce human error. If a switch cannot accept the change, an alert should be raised and no switches should accept the change until this is resolved.

The SDN control and state is “in the network” running inside the switches themselves to avoid the cost, latency penalty, and reconvergence risk of a remotely located controller keeping state. Additionally, a controllerless solution avoids the reliability risk using the out-of-band (OOB) management channel and corresponding single OOB port on the switch.

Pervasive telemetry must be throughout the fabric down to the flow level so problems anywhere in the fabric can be quickly identified rather than a time-consuming switch-by-switch investigation.

This functionality should be available on open networking switches so white box economics can be achieved as many switches need to be deployed in many locations.

This approach goes beyond simple point-to-point connectivity to enable seamless unification of multiple data centers. It supports high-performance and scale-out architectures that allow multiple data centers to work together more eff­iciently — dramatically simplifying operations, reducing the potential for human error, and increasing resiliency and security.

Next generation DCIF has tangible benefits for enterprises on their digital transformation journey including:

Disaster recovery: Most data center owners have a redundancy strategy and at least two data center locations for disaster recovery. Data center fabrics can make this failover across two or more sites simpler by leveraging technologies like distributed routing with anycast gateway functionality.

Streamlined data center migration/consolidation: For workloads that are still on premise, a high capacity DCI fabric can dramatically streamline the movement of workload from a current data center location to a new one.

Simplify operations: The opportunity here is to make 40 switches look like one switch, with all 40 being controlled from any switch via CLI or from any switch via a REST API with full CLI parity. Well implemented data center fabrics with next-generation software defined networking offer this promise.

Preparing for edge compute: With the proliferation of IoT and new applications that require ultra-low latency, data centers are going to need to become even more distributed. Organizations are now beginning to deploy smaller edge compute data centers — either on-premise or with a colocation provider. DCIF can help to simplify this implementation.

The time is now to unify your data center infrastructure. Modern DCIF provides innumerable benefits and puts companies in a position to deliver elastic, scalable, and high performance digital experiences to employees, customers, and partners. If any of these issues are impacting a company’s data centers or business, now is the time to consider a unification strategy.