UPS and Power Protection for ISP Sites

Of all the ways an ISP goes dark, power is the most common and the most total. A router can fail and reroute around itself; a single fiber can break and leave most customers untouched. But when a site loses power, every customer behind it loses service in the same instant, and they all call at once. Power is the failure mode with the widest blast radius and the least subtlety, and it is also, frustratingly, the least glamorous part of the build — the part that gets value-engineered out of the budget first and remembered only during the outage it was supposed to prevent.

That is a shame, because power protection is some of the cheapest reliability money can buy. A modest investment in batteries, a generator and basic electrical hygiene removes an entire category of total-site outages. The catch is that power protection is not a product you install once and forget; it is a small system that must be sized correctly, prioritised sensibly and — above all — tested relentlessly, because every part of it degrades silently until the day you need it. This guide covers how to think about all three.

The layers of power protection

Power protection is built in layers, each covering a different duration and failure mode. Used together they form a chain from a momentary flicker to a multi-hour blackout:

• UPS (battery): the first line, and the only one fast enough to matter. A UPS bridges short outages and voltage sags instantly, with no gap, and buys the seconds or minutes a generator needs to start. Just as importantly, a good UPS conditions power — smoothing the spikes, sags and noise on the incoming mains that quietly shorten the life of every power supply behind it.
• Generator: the second line, for outages longer than the battery can cover. Once a blackout outlasts your runtime, only a generator keeps the site alive. It is essential at core and major aggregation sites where the downtime would be widest and longest.
• Dual feeds and redundant supplies: the deepest line. Critical gear with two power supplies, each fed from an independent source, survives a single failure on one side without the customer ever noticing. This is the same single-fault-survival thinking that runs through good network redundancy design, applied to electricity.

Size it honestly: load and runtime

A UPS is defined by two numbers, and getting either wrong means it fails at the exact moment it was bought for. The first is load — how much power your equipment actually draws, not a guess and not the nameplate maximum. The second is runtime — how long the battery must hold that load, which depends entirely on what the next layer is. If a generator backs the site, runtime only needs to cover the start-and-transfer gap with margin. If there is no generator, runtime is your entire defence and must cover a realistic outage.

Undersize the load rating and the UPS overloads and drops the moment a real outage hits. Undersize the runtime and the battery dies mid-outage, which is arguably worse than no UPS at all because you planned around protection that was not there. Size for the measured load with headroom for growth, size runtime to your actual strategy, and then re-check both periodically — because sites grow quietly. Engineers add a switch here and a server there, and a UPS that comfortably covered the original load is silently overcommitted a year later.

Prioritise by blast radius

You cannot gold-plate every cabinet, and you should not try. Spend where an outage hurts the most people, which means matching the depth of protection to the number of subscribers behind each site. The principle is simple: the wider the blast radius, the deeper the protection.

A POP aggregating thousands of subscribers deserves far more protection than a cabinet serving a handful, and budgeting that way concentrates spend where it removes the most customer-minutes of downtime per dollar.

The part everyone skips: testing

Here is the uncomfortable truth about power protection: the parts you do not test are not protection, they are decoration. Batteries age and lose capacity invisibly — a battery string rated for thirty minutes when new may deliver only a few minutes after a few years, and nothing on the front panel warns you until it fails under load. A generator that has not been started in months is just expensive scenery; fuel goes stale, batteries flatten, and starters seize, so the one time you need it, it cranks and dies.

The discipline is straightforward and non-negotiable. Test the transfer to generator under real load on a schedule, not just an idle start. Check battery health regularly and replace strings before they fall below your runtime target rather than after. Periodically confirm that actual runtime still matches what you assumed when you sized it. The worst possible moment to discover a dead battery or a generator that will not start is during the outage it existed to cover — and that is exactly when untested power protection always fails.

The transfer gap: where protection actually breaks

The most overlooked failure in a layered power design is not a dead battery or a generator that will not start — it is the handoff between layers. When mains fails, the UPS must carry the load instantly while the generator detects the outage, cranks, stabilises and takes over through a transfer switch. Every one of those steps is a place the chain can break. If the UPS runtime is shorter than the generator's real start-and-stabilise time, the battery empties before the generator is ready and the site goes dark despite having both. If the transfer switch misbehaves, the generator runs while the load stays on dead mains. These are not exotic faults; they are the ordinary way that sites with full protection still go down.

The defence is to test the whole chain end to end under real load, not each piece in isolation. A generator that starts fine on a manual test can still fail to pick up the load through the transfer switch in an actual outage, and a UPS that holds for its rated time on paper can fall short once its batteries have aged. Time the real gap between mains loss and stable generator power, confirm your battery runtime comfortably exceeds it with margin, and rehearse the transfer often enough that you trust it. Protection is only as strong as its weakest handoff, and the handoff is exactly the part nobody tests until the night it matters.

Don't forget the basics: earthing, surge and monitoring

Two foundations sit underneath all of this. Proper earthing and grounding, together with surge protection, guard against both the slow degradation and the sudden death that electrical faults and lightning inflict on equipment — damage that batteries and generators do nothing to prevent. Skimping here means losing gear to problems your UPS was never designed to stop. And finally, monitor your power as seriously as you monitor your network: alert on mains loss, low battery and generator status so that you learn about a power problem from a dashboard the moment it starts, not from a rising wave of customer calls an hour later. Monitoring also surfaces the slow drift — the battery that is degrading, the feed that flickers nightly — before it becomes an outage.

The takeaway

Power is unglamorous, and it is where a surprising amount of uptime is quietly won or lost. Layer UPS, generator and redundant feeds; size honestly for both load and runtime; concentrate protection where the blast radius is widest; and test everything relentlessly, because every component degrades in silence. Get the electrical foundation right and the network on top of it — and the management and monitoring layer ISP Digital provides to watch it — has a stable base to stand on instead of a single point of total failure.