Planning a modernization project is the easy part, relatively speaking. Everyone can build a nice roadmap, rank risks, get budget approved. The part that actually determines whether the project goes well or turns into six months of pain is execution — the actual mechanics of swapping out equipment that's carrying live traffic, without breaking the business that depends on it staying up while you do it. That's the part most modernization conversations gloss over, and it's the part I want to actually dig into here.
I'll say this bluntly: a lot of modernization projects fail not because the plan was wrong, but because the execution treated a live, business-critical network like it was a lab environment you could just swap pieces on and see what happens. It isn't. Every cutover step needs to assume something might go wrong, because eventually, on some project, something will.
Start With a Change Window That Actually Reflects Your Business, Not a Generic "Weekend"
The default instinct is always "we'll do it over a weekend." For a lot of businesses that's genuinely fine. For others — retail during a peak sales period, professional services during month-end close, anyone with real weekend traffic — a generic weekend window is actually one of the worst possible choices, and nobody stopped to check the calendar against the actual business before picking it.
Know your own traffic patterns before you pick a window. If Sunday night into Monday morning is genuinely your quietest stretch, that's your window, even if it's less convenient for the team doing the work than a Saturday would be. The business's actual tolerance for disruption should drive the schedule, not the team's preference for a normal weekend off.
Stage Everything You Can Before Touching Live Traffic
This sounds obvious and gets skipped constantly under time pressure. New equipment should be configured, tested, and validated in a staging environment — or at minimum, configured and burned in on the bench — before it ever touches production traffic. Every setting that can be pre-configured offline should be pre-configured offline, so the actual cutover window is spent swapping and verifying, not troubleshooting a configuration mistake live while the business is watching a dashboard turn red.
We've seen projects lose hours of a tight change window to a typo in a VLAN configuration that could've been caught in five minutes of staging review. That's not a technical failure. That's a process failure — skipping staging because the window felt tight enough already, which is exactly backwards, since staging is what makes the window survivable in the first place.
Sequence the Cutover by Dependency, Not by What's Easiest to Reach
The instinct under time pressure is to start with whatever's simplest — swap the easy switch first, work up to the harder stuff. That's usually the wrong order. Map out what actually depends on what, and sequence around genuine dependencies: core infrastructure before edge devices that depend on it being up, anything supporting critical applications handled with extra care and extra validation before moving on to lower-priority systems.
Getting this sequence wrong doesn't just risk one failed step — it risks a cascading set of problems where fixing step three requires undoing step five, which you already completed, because step five turned out to depend on something in step three that changed.
Have a Genuine, Specific Rollback Plan for Every Single Step
"We'll roll back if something goes wrong" is not a plan. It's a hope wearing a plan's clothing. A real rollback plan answers, for every individual step in the cutover: what exactly does undoing this step actually involve, how long does it realistically take, and who's specifically authorized to make that call in the moment, under pressure, without waiting for a committee.
If you can't answer those three questions for a given step, you haven't actually planned that step — you've planned the parts of it that go well. The parts that go badly are exactly where a real plan earns its cost.
Test as You Go, Not Just at the End
Waiting until the entire cutover is complete to start testing means any problem found late could be sitting anywhere in a long chain of changes, and untangling which specific step introduced it eats real time you don't have in a tight window. Testing incrementally — confirming each major step actually works before moving to the next — catches problems close to their source, while it's still obvious which change caused them.
This takes real discipline under time pressure, because testing feels like it's slowing you down when the clock's already ticking. It's actually the opposite. A problem caught immediately after the step that caused it is a five-minute fix. The same problem discovered three steps later, buried under everything that happened since, can eat the rest of your window.
Overcommunicate During the Window, Even When Nothing's Wrong
Silence during a change window makes people nervous, and nervous people start assuming the worst, escalating, asking questions that pull attention away from the actual work. A short, regular update — "still on track, moving to phase two now" — costs almost nothing and keeps stakeholders calm even when there's genuinely nothing new to report yet.
This matters more than it sounds like it should. The alternative to a boring, regular update isn't silence being interpreted as "everything's fine." It's silence being interpreted as "something's wrong and nobody's telling us," which is a much worse position to be defending from if you do eventually need to report an actual delay.
Don't Decommission the Old Equipment the Moment the New Stuff Looks Fine
The instinct to tear out old hardware quickly — clean up the rack, stop paying for equipment you don't need anymore — is understandable and, almost every time, premature. Some issues don't surface in the first hour. They surface during the first real Monday morning traffic spike, or the first month-end batch job, or the first time a specific edge case gets hit that nobody happened to test during the cutover window itself.
Keep old equipment available, even just powered off and ready to reconnect, for at least a couple of weeks past cutover. It's a small ongoing cost against a genuinely large one if something surfaces later and the fallback option's already been physically removed from the building.
Document What Actually Happened, Not Just What Was Planned
The plan and the actual execution rarely match perfectly — something always comes up, some step takes longer than expected, some workaround gets improvised in the moment. That gap between plan and reality is exactly what the next modernization project needs to learn from, and it only gets captured if someone actually writes it down while it's fresh, not reconstructed from memory months later when the next project kicks off.
What Genuinely Careful Execution Looks Like
Pulled together, this generally means:
- A change window chosen around your actual business traffic patterns, not a generic default
- Everything staged and validated before it touches live traffic, so the window is for execution, not troubleshooting
- Cutover sequenced by real dependency, not by whatever's easiest to reach first
- A specific, answerable rollback plan for every individual step, not a vague overall intention
- Incremental testing throughout, catching problems close to their source
- Regular, boring communication during the window, even when there's nothing new to say
- Old equipment kept available for weeks past cutover, not torn out the moment things look fine
- Honest documentation of what actually happened, not just what the plan said would happen
The Actual Point
Nobody executes a network modernization flawlessly against the original plan. That's not a controversial thing to say — ask anyone who's actually run one. The projects that go well aren't the ones that avoided every surprise. They're the ones with rollback plans specific enough to actually use, testing thorough enough to catch problems close to where they started, and enough discipline in the execution that the surprises which do show up stay small and manageable instead of cascading into something much worse.
Good planning gets you a solid roadmap. Good execution is what actually gets you through the weekend without a 3 a.m. phone call.
Top comments (0)