Normally, data center networks are updated when new applications or servers are installed in the infrastructure. But independent of new server and application infrastructure forcing an update, there are other areas to consider. Three questions to ask when assessing if you need to update your network are:
- How are server speeds dictating network design?
- Are your network features out of date?
- Is your operational workflow inefficient?
How are server speeds dictating networking design?
Network device selection typically starts with understanding how the server Network Interface Cards (NICs) are configured. In the past, server NICs at 10 gigabits/second (10G) were considered the norm. But over the past 5 years we’ve seen a real growth in server computing power. In the accelerated computing world we tend to see 25 to 100G network speeds as the new norm for servers, with the latest servers able to use even 200G NICs.
With higher NIC speeds, the top-of-rack (leaf) switch needs to be upgraded. Failure to update your legacy core (spine) switches will cause oversubscription ratios to move unfavorably, introducing excess congestion and unpredictable latency. If you’re upgrading the leaf switches, you’ll need to upgrade the spine switches as well. Maintaining the same oversubscription ratio should be the goal.
Are your network features out of date?
In addition to the hardware, it may be time to upgrade your network operating system (NOS), especially if you’re using legacy network features and protocols that are hindering your network through inefficient management and packet forwarding. Legacy networks were often built using layer 2 infrastructure everywhere. The entire network would be on a single broadcast domain, and solutions like MLAG would provide one layer of redundancy for every link.
Updating the network to leverage solutions like Layer 3 to the host (also known as host based networking) or VXLAN+EVPN overlays, helps alleviate issues caused by inefficient broadcast domains. Host based networking and overlay networks enable better traffic management, easier provisioning, and more granular security. Anytime is a good time to update your network if it’s relying upon older technologies that aren’t modernized.
Is your operational workflow inefficient?
Beyond the infrastructure hardware and software, an area of optimization that benefits from modern updates is the operational workflow. Classic networks tend to have network admins log in to each switch, router or firewall individually. Configurations are applied unique to the node, and are backed up using a plain text file and stored locally on the network admin’s computer. These workflows tend to be error prone and can lead to typos or inefficiencies such as lost backups and slow maintenance windows. Configuration errors can open security gaps that a cyber adversary could exploit.
Modern networks leverage more advanced tooling and technologies that solve most of those problems.
Updating your networking can add benefits, including:
- Infrastructure as code, so your configurations are centralized.
- Automation, which allows managing and updating multiple nodes at the same time.
- Continuous integration that systematically validates configuration and design prior to deployment.
- Network simulation through digital twins. This helps predict how the network will behave and tie together all the elements from network DevOps to automation to continuous integration.
These optimized workflows reduce maintenance window times and errors in deployments, reducing the risk of security gaps while saving time and effort.
There are many reasons to update a network. It’s important to take stock of your network’s hardware, software and operational efficiencies and look at the big picture. With this information you can determine if updating your network will result in faster throughput, more productivity, and lower ownership costs.