How to reduce networking costs without performance loss

Cost drivers in enterprise networking

IT teams inadvertently drive network costs by improperly rightsizing, increasing technical debt and focusing on manual work rather than automating.

Rightsizing

Rightsizing connectivity applies throughout the network, from the server or desktop edge to the WAN and cloud networks. Many enterprise network teams build networks with equipment intended to last multiple years. This attempt at futureproofing hinders network development and increases costs, however.

When organizations buy equipment they think they'll use for multiple years, they end up overprovisioning the network and spending more on infrastructure that soon becomes outdated. Not only does this limit network upgrades, but it also causes organizations to pay too much to buy more than they need.

Technical debt

The problem of technical debt in the network manifests itself in both deferred work and extended lifespans:

• Deferred work. Backlogs of OS and firmware updates, postponed configuration maintenance, incomplete equipment inventories and other operational issues. This often leads to more downtime, longer restoration times and increased cybersecurity risks. 
• Extending lifespans. Operating hardware past its end of support increases costs for the vendor and secondary-market support. It also increases costs due to outages as older systems become more prone to failure. Older hardware is also generally less power-efficient, costing more to keep.

Manual work

The continued reliance on manual network management drives excess costs. Manual management is both slower and more error-prone than automatic processes. As network salaries continue to rise, downtime costs increase. At the same time, the costs of automation tools are decreasing. In a time when it becomes harder to hire experienced network professionals, organizations should feel incentivized to consider fully automating network management to offset the skills gap and lower costs.

Reduce networking costs without degradation

Enterprise network leaders need to attack excess costs on all fronts.

Proper rightsizing

When network teams design or upgrade their networks, they should intentionally create or retain excess capacity solely to enhance network resilience. To address overprovisioning, leadership must encourage network engineers to forgo old rules of thumb. For example, it's better to make design decisions based on network traffic data rather than outdated guidelines, such as ensuring link capacity exceeds the highest estimated load by 40%.

This can lead to significantly lower edge and spine connectivity costs within a campus or data center network. Coupled with the use of scalable connectivity strategies on an SD-WAN and flexible cloud-exchange-based connections to cloud providers instead of inflexible direct connects, it can lead to significant reductions in WANs as well.

Reduce technical debt

Teams can focus on replacing the oldest tiers of equipment: First, out-of-service items and then items near that status. It's important to implement strong lifecycle management, based on accurate inventory and configuration management, from the moment new gear rolls in.

New network discovery and management tools can sometimes pay for themselves in this way, leaving aside savings on downtime and reduced cyber risk. Eliminating the oldest and least reliable equipment will drive short-term Capex increases but should reduce ongoing Opex enough to outweigh them.

Automation in cost reduction

Network teams should select and standardize on a limited set of automation tools and approaches and apply them to every aspect of operations. The key to savings is automating the remaining manual operations that are most time-consuming. This is true whether rolling an in-house tool, working with a third-party option, using a declarative or imperative style or fully embracing an infrastructure-as-code discipline.

That doesn't always mean automating the single biggest time sink, however, because it's important to recognize the complexity of the task teams want to automate. Addressing a few simple problems in one week could free up as much staff time as it would take to fix one complex problem that requires a full month.

Teams can boost their chances of automation success with the following:

• Deploy consolidated and virtualized branch stacks.
• Use white box switches running centralized SDN software.
• Adopt carrier and cloud services with APIs that enable programmatic access to features, such as route management and dynamic bandwidth increases and decreases.

Visibility and monitoring lower costs

Network teams can't manage what they can't see. To know where to focus cost-reduction efforts, the network team needs as much visibility into the network and its operations as possible, including infrastructure, services and staff time.

Modern management tools include traditional network management suites, SD-WAN platforms and cloud cost analysis tools.

Teams can use these tools to see the following:

• Capacity exceeding demand.
• Networking services fee changes that affect the budget.
• Staff troubleshooting and management time exceeding the average.
• Low performance and reliability areas.
• Systems past end of support or end of life.

Teams can also make data-driven projections of growth and traffic changes, as well as predict service impairments from potential equipment failures.

Cost control in a modern network

The amount of money and time spent keeping the network up can be better spent toward improving the network. Rightsizing, adopting widespread automation and drastically reducing technical debt can result in lower operational costs and lower barriers to scaling and introducing new IT services.

Getting there requires organizations to make focused, short-term investments in new generations of tools, services and equipment to garner ongoing, long-term savings and service improvements. It also means investing in the professionals who run the network, helping them gain and polish skills needed for automation and continually adjust the network to meet evolving needs.

John Burke is CTO and a research analyst at Nemertes Research. Burke joined Nemertes in 2005 with nearly two decades of technology experience. He has worked at all levels of IT, including as an end-user support specialist, programmer, system administrator