Network Engineering, November 2020

Where Network Engineers Actually Work

Job titles in this field carry almost no information. The same title covers a role configuring switches for one site and a role designing a backbone across a continent.

The type of organisation is a much better predictor of what the work is, what is learned, and where it leads.

The six categories

TypeThe work is mostlyWhat is learned
EnterpriseOperating a network built by someone elseBreadth, and the politics of change
Service providerScale, routing policy, customer isolationDepth in routing, and real consequence
HyperscalerAutomation over configurationSoftware practice applied to networks
VendorDesign, support, or developmentOne product line very thoroughly
IntegratorMany networks, briefly, to a deadlineSpeed and variety, rarely depth
Managed serviceMany customers, ongoing, at volumeProcess, and diagnosis under pressure

None of these is better. They select for different temperaments, and someone who thrives in one is frequently unhappy in another for reasons that have nothing to do with ability.

Scale changes the problem, not the amount of it

A larger network is not the same work done more times. Beyond a certain size, manual configuration stops being viable, and the job changes from operating devices to operating a system that operates devices.

That transition is the most significant fork in the career. Engineers who make it are writing and reviewing automation, and those who do not are increasingly maintaining the parts that have not been automated yet.

Where the interesting problems concentrate

Provider and hyperscaler environments encounter protocol behaviour that enterprises never see, because they run at a scale where the corner cases actually occur. Route reflection design, convergence tuning, and the failure modes of large routing domains are daily concerns rather than certification material.

Enterprises encounter a different difficulty, which is heterogeneity. Equipment from several vendors, several generations, and several acquisitions, operated by a team without authority over most of it, is a harder organisational problem than a technical one.

What actually gets people hired

Certification opens the first door and stops mattering quickly. Beyond the first role, the questions concern what was built, what broke, and what was done about it.

Demonstrable automation ability is the strongest differentiator currently available, because the supply of engineers who can configure a device far exceeds the supply who can describe a network as code and test a change before applying it.

Choosing deliberately

The useful question when evaluating a role is not the title or the technology list. It is what proportion of the time is spent operating, designing, and building, because those three produce very different careers from the same starting point.

Asking directly how the last significant change was made, and how it was tested, reveals more about the environment than any description of the stack.

Note: the most common career mistake is staying in an environment where the network never changes. Stability is comfortable and produces very little learning, and the gap compounds quietly over several years.