Data-center network automation: Its pitfalls and how to avoid them

(Enterprise Management Associates has published “The Future of Data Center Network Automation” based on a survey of enterprises, cloud providers and network service providers. This article by EMA Vice President of Research Networking Shamus McGillicuddy details some of its major findings.)

More than 86% of companies expect their budgets for data-center network automation to increase over the next two years, and with spending ramping up, network teams need to plan carefully. Network automation is notoriously difficult to implement due to the complexity of networks in general, and only 23% of the people surveyed were fully confident in their data-center network-automation strategies.

Asked to identify their biggest challenges across three stages of their automation initiatives—planning and evaluation, technology implementation, and use of automation—the common thread was the challenge of integrating and contextualizing network automation with end-to-end digital infrastructure.

Planning and evaluation challenges

When planning and evaluating data-center network automation, technology organizations most often struggle to understand how it will interact with other tools and management systems (39%). Network infrastructure and operations teams often have a multitude of tools for managing networks, including multiple tools for data-center network automation. On top of that, other teams—including systems, storage, security, applications, and DevOps—are managing aspects of data-center operations with their own toolsets. The network team needs to think about how their automation toolchain will interact with these others. 

Many organizations (37%) say they are struggling significantly with planning budgets and understanding the costs associated with data-center network automation. EMA believes that the complexity of an automation initiative is to blame for the lack of clarity around costs. Most of the organizations surveyed are both purchasing commercial automation tools and developing home-grown automation software. It is difficult to project how much it will cost to implement and maintain this constellation of tools.

Implementation challenges

When implementing data-center network automation, 44% of organizations struggle significantly with infrastructure issues. Network devices and other components in the data center have legacy issues that make automation more difficult. For instance, older equipment might lack APIs, which forces the automation team to implement a solution that pushes changes to the network via command-line interface (CLI) scripts.

Even more problematic, networks often have multiple versions of vendors’ network operating systems (NOS) in production, and each NOS version might have subtle differences in CLI syntax, creating even more complexity. Network engineers have told EMA that the APIs on some of their more modern equipment are limited in scope, functionality, and quality, which creates even more infrastructure issues.

Nearly half (44%) of organizations are struggling to integrate and contextualize their automation pipelines with overall application service delivery. This echoes the planning and evaluation issue of understanding how automation will interact with other systems. Network teams struggle to build an automation solution that allows them to push network changes with a full understanding of how those changes will affect the behavior, security, and performance of applications.

Automation use challenges

When using automation, data authority and quality issues are a major problem for 42% of organizations. Data is the lifeblood of network automation. Engineers need data about the state of the network, such as device metrics and traffic flows, to understand what automated changes are required. They also need data about network intent, such as configuration standards and security policies, to implement acceptable changes. Unfortunately, many organizations struggle to create a reliable repository of data to ensure effective automation.

As a network architect with a $50 billion consulting company described the issue, “You can push configuration to thousands of boxes, but if that configuration is wrong, you’ve created one hell of an outage.” Users of automaton, like planners and implementers, also worry that network changes will impact application behavior and performance, and 40% of IT organizations identified this as major issue. This struggle to contextualize and integrate network automation with the rest of digital operations was a consistent through-line in EMA’s research.

What to do

Best-in-class organizations focus on improving network compliance with automation, which will force a rigorous approach to establishing a repository of network intent data. Best-in-class organizations also look for automation tools with change analysis and modeling features, which can help them understand how network changes will impact overall digital operations.