New technologies like 5G, AI, and IoT, are driving companies to reconsider the way their data centers are structured. But from the smallest companies all the way up to the largest cloud providers, we’ve seen that service uptime remains a challenge for even the largest and most sophisticated companies – some citing that an outage can cost them more than $1 million.

The frustration is all too real: How can you be expected to operate (much less innovate) when your network is stuck in the past? The network must lead the IT modernization charge.

Modernizing the data center network is a rewarding endeavour, but due to the hurdles that can be encountered, some hesitate to make the transition. However, not doing so means staying in a traditional, slow and outdated system. I’d say the rewards outweigh the potential risks. From my experience, if folks follow a set of principles, the move to modern becomes more efficient and less of a challenge. Here are the top five guiding principles that I’ve found to be the best to follow when modernizing the data center network

1. Use standards-based protocols and services for the network architecture: Open-source technology has gained popularity over the last couple of years for good reason: Standards-based protocols promote interoperability, competition and innovation, unlike proprietary protocols that require specialized engineers, limit interoperability and lock organizations into specific designs.

2. Ensure that the network is serviceable without downtime: To get into the nitty-gritty of it, all compute nodes need to be dual-connected to redundant upstream Leaf switches in order to keep outages from occurring. Leaf switches should have redundant peer-link connections between each other and to each Spine switch. Equal cost multipathing — a useful failover tool and helpful with load balancing — ensures that all paths are active and forwarding. This way, if a Leaf or Spine switch is to be inserted or removed, the production of traffic won’t be affected.

3. Have a network architecture in place that promotes automation: When designing or monitoring a new network, it’s important to verify that it’s running as intended and follows network and security policies in place. Making configuration changes manually tends to be extremely time-consuming and can be fraught with human error. But by automating tasks instead, the network can be enabled to be more consumable, self-healing, and less complicated to audit.

With familiar Linux APIs, DevOps engineers are enabled to incorporate the network into automation engines without the difficulty of having to handle a number of APIs that are vendor-specific. No matter the underlying hardware, having the same network operating system (NOS) opens the door for simplified network automation.

4. Set up the network to be consumable: Having the capability to empower administrators or customers with self-deployable networks should be a key consideration with new network designs, whether the data center is private and serving a single organization or built for a busy IaaS platform. Consumable self-service networks tie into the concept of automation and today, creating networks in the public cloud is a fundamental feature everyone expects.

The infrastructure of the cloud-based data center allows organizations and their IT to consume additional compute and storage cycles on an as-needed basis. Due to native Linux models and APIs, a Linux NOS is ideal for orchestration solutions. Deployments that harness EVPN with automation facilitate the deployment of new networks while simultaneously enabling customers to build their own on the fly, without the intervention of network engineers.

5. Construct the network to be scalable: With equal-cost multipathing of 128 links, Leaf-Spine pods can become massive, which is why a Leaf-Spine Clos architecture is ideal for data centers. Additional pods can be added to grow horizontally or new tiers can be added to grow vertically, interconnecting an infinite amount of pods. A disaggregated model enables data center admins to swap hardware modularity, automating the NOS and network provisioning with ONIE as well as offering flexibility, should port-density or port-speeds in certain spots become insufficient.

Transforming the network can be a challenge, but it’s one that organizations have to overcome if they want to be able to meet the new needs that come with the growth of cloud and innovative technologies. These principles, while not all-encompassing, will enable organizations to achieve a modernized data center network that is built to scale, is flexible, and can meet the needs of business today.

Follow these five guiding principles through from project inception to network deployment, and you’ll succeed in designing and deploying a new modernized data center network.