Maximize Data Center Network Automation with the Right NOS

Summary

Automation is more important than ever in modern data center networks. It’s the go-to solution that every organization turns to when its IT staff gets overwhelmed, contending with increasingly complex operational environments and too many high-performance applications.

With so many organizations now relying on network automation, why has it never realized its true potential?

It’s not a stretch to say that most organizations believe automation will boost their productivity and simplify operations. Yet many organizations appear reluctant to execute automation at full scale or across many linked tasks, wary that it could break the network or amplify its shortcomings. This hesitation is reflected in a study by Enterprise Management Associates, which reveals that most organizations have automated less than half of their data center network management tasks.

How the network makes automation better

In order to get full value from network automation, organizations must overcome these barriers to adoption. Typically, the main challenges are not solely on their automation platforms, but in their networks. Here, there are two key challenges to address.
 
First, an automation platform is only as good as the data it receives and processes from the network. While many automation platforms are sophisticated enough to meet the minimum automation demands, they need to understand the network and its state in near-real time. But, their effectiveness is limited by the granularity and expediency of the network data and how easily they can access it.
 
Second, networks really are fragile. Using automation to accelerate or scale network operations by an order of magnitude could expose and amplify this fragility. It’s natural to worry, but operators need solutions that will convince them that their networks can handle automation.
 
So, how do organizations make their data center networks more readily available to, and consumable by, their chosen automation platforms? And how do they make sure that their automation ambitions won’t be undone by a hidden network issue?
 
The answer is the same in both cases: Choose the right network operating system (NOS).

The right NOS unleashes network automation

The NOS is the gateway to the network, and its construction determines how available and consumable the network really is. To support full-scale automation without limits, compromises, or worries, a NOS must ensure that the automation platform has fast and easy access to its current configuration and state data across all subprocesses and services.
 
An ideal NOS will address this need with a model-driven management architecture and a microservices approach that models each service using a standard data modeling language such as Yet Another Network Gateway (YANG). It will also provide the automation platform with access to network data through streaming telemetry collected through one application programming interface (API) such as the gRPC Network Management Interface (gNMI). Many legacy NOSs are not built uniformly across the entire system and to access all the network data needed, often many APIs are required (e.g. CLI, SNMP, Netconf, etc.). Using multiple APIs for this purpose will slow down the platform and make automation more difficult.
 
The NOS must also provide network state data as part of its telemetry stream. This is essential for ensuring that the automation platform can always evaluate important state information before executing automation tasks. Having access to simple information, such as whether a port is available, or complex information, such as whether Border Gateway Protocol (BGP) neighbor adjacencies are established, could be equally essential for a specific automation. A NOS that provides this data can ensure that the automation platform always has the information it needs.

Pre-automation checks identify network fragilities

With this foundation in place, the NOS can allow an organization to use its automation platform to conduct checks to verify that a complex automation task won’t break the network, much like pre-flight checks identify issues with airplanes. For instance, an organization could use intent templates to pre-program the platform to conduct automation or network health checks to identify problems before a task is executed.
 
Checks could also be used to identify potential fragilities at every key stage of the automation lifecycle, from Day 0 design to Day 1 deployment and Day 2+ operations. These can be simple checks, such as verifying that all required ports are operating, or complex, such as confirming the state of the Routing Information Base (RIB) or verifying that all congestion levels are within acceptable thresholds.
 
With a NOS built on a model-driven management architecture, an organization can make pre-automation checks more granular and precise, and conduct them across the network in a faster and more consistent way.

It’s time to see what full-scale automation can do

Organizations no longer have to keep their network automation ambitions in check. By choosing a NOS that provides granular, easy-to-consume network data across all services and supports pre-automation checks, they can confidently begin to realize the full potential of data center automation.