Data centre power redundancy means no single path and no single point of failure between the utility and your rack. Power moves from Toronto Hydro through an automatic transfer switch and panel to a UPS, then to managed PDUs, then to your equipment. The UPS never stops feeding your rack. If the grid drops, it carries the load on battery while the transfer switch brings the backup generator online. N+1 means every critical system runs at least one unit beyond what the load requires, so a single failure or a maintenance window never reaches your servers.
If you are choosing a colocation provider, one question sits underneath every power spec on the tour. When the grid drops, or when somebody has to service a UPS on a Tuesday afternoon, does your equipment find out? A facility engineered properly makes the answer no, every time, and can show you why.
Power redundancy comes down to a single rule: no single path and no single point of failure anywhere between the utility and your rack. Each link in the chain below protects against a different failure, and each one is a place a cheaper facility can quietly economize.
What does the data centre power chain actually look like, piece by piece?
Electricity entering a data centre passes through several distinct stages before it reaches your equipment. Each stage exists to solve a different failure mode, and none of them is optional.
Toronto Hydro, the utility grid
The starting point, and the least reliable link in the chain. Toronto Hydro power fluctuates, sags, spikes and occasionally fails outright. No serious data centre treats utility power as sufficient on its own.
The automatic transfer switch and panel
The automatic transfer switch watches the incoming utility feed and, the moment it drops, calls in the backup generator and switches the panel over to generator power. There is no human in the loop and no phone call to make. The panel then distributes that power onward to the UPS, whether it is currently sourced from the grid or from the generator.
The UPS, or uninterruptible power supply
A UPS sits between the panel and your equipment, holding a battery charge that takes over the instant incoming power drops, with no perceptible interruption. It is the one component in the chain that is always feeding your rack, whatever is happening upstream.
Its job is not to run the building for hours. It is to buy the couple of minutes the generator needs to start and take over the panel. Without a UPS, a grid blip measured in milliseconds would be enough to crash every server in the building.
Backup generators
Diesel generators are the long-duration backup. A UPS measures its runtime in minutes; a generator measures its runtime in weeks. At 151 Front Street West the generator plant holds one to two weeks of runtime on the fuel kept on site, and beyond that it runs for as long as resupply keeps arriving.
They start automatically on grid failure, and they never connect straight to your rack. They feed the panel, which feeds the UPS, which stays in the loop powering your equipment the entire time. How much runtime a facility holds on site, and how many generator units it runs, is one of the clearest signals of how seriously it takes an extended outage.
Managed PDUs, or power distribution units
Once power has been conditioned by the UPS, managed PDUs distribute it out to individual racks and circuits, the way a breaker panel distributes power around a house. In a data centre they also monitor load per circuit, which is how a facility knows a rack is approaching its ceiling before a breaker tells it.
That ceiling is a real number, and it varies by suite. Ours is 15kW per rack in TOR1 and 30kW per rack in TOR2. Per-outlet metering on most managed rack PDUs then lets you see exactly what your own equipment is drawing against it, in near real time, without asking anybody.
| Component | What it protects against | How long it holds |
| Automatic transfer switch | A human being slow. Moves load between sources with nobody in the loop | Instant, and continuous |
| UPS | The instant of grid failure, and every sag and spike short of one | Minutes, enough to bridge to generator |
| Backup generators | An extended grid outage | One to two weeks on site, then as long as fuel resupply continues |
| Managed PDUs | Silent overload. Distributes and meters power per circuit and per rack | Continuous monitoring |
What is N+1 redundancy, and why is it the number that actually matters?
N is however many units it takes to carry the full load. N+1 means one additional unit exists beyond that, on every critical system, so any single unit can fail or be pulled for maintenance without the load ever noticing. N+1 is what makes maintenance possible without downtime, which is why it matters more than any single component on a spec sheet.
Say a facility’s UPS load requires three units running at full capacity to support everything connected. Running exactly three, with zero spares, is N redundancy, which is to say no redundancy at all. One unit failing, or one unit needing scheduled service, immediately puts the facility at risk. N+1 means running four units for that same load, so any single one can go down, planned or unplanned, and the other three absorb it without anyone downstream noticing.
The same arithmetic applies to generators, cooling and distribution. Redundancy is not a property of one machine. It is a property of every critical system in the chain at once, which is why a facility that quotes you a UPS model without quoting you a topology has not answered the question.
N+1 versus 2N: what does the difference actually buy you?
N+1 adds one spare unit to a shared system. 2N duplicates the entire system, so there are two complete, independent sets of everything and either one can carry the full load alone. 2N costs roughly twice as much to build and to run, and it protects against a whole-path failure rather than a single-unit failure.
For most colocation workloads, N+1 across power, cooling and network, with independent distribution paths, is the point where marginal risk reduction stops being worth marginal cost. 2N is generally a decision driven by a specific regulatory or contractual requirement rather than by engineering.
Where does a cheaper facility actually cut corners?
The five places worth checking: N+1 claimed on one system but not on every critical system, redundant units sitting on a single shared distribution path, generator capacity sized for the IT load but not for the cooling that keeps the room habitable, redundancy that has never been load-tested under real load, and a Tier III building presented as a Tier III deployment.
None of these show up on a spec sheet, and all five are answerable on a tour. Ask which systems are N+1 and which are only N. A facility that runs N+1 UPS but N cooling has a single point of failure that will take the room down slightly more slowly, which is not the same as not taking it down.
Ask whether the redundant units share a distribution path, because four UPS units behind one path fail together. Ask whether the generators carry cooling as well as IT load, because a room that stays powered and then overheats has failed just as completely. Ask when the transfer was last load-tested, because untested redundancy is a theory. And ask whether the Tier III designation belongs to the building or to the deployment you are buying, a distinction we come back to at the end of this piece.
What is concurrent maintainability, and why does Tier III require it?
Concurrent maintainability means every piece of power and cooling infrastructure can be repaired, replaced or upgraded without removing power from the load it serves. It is the core requirement of the Uptime Institute’s Tier III standard, and N+1 is the mechanism that makes it physically possible.
This is the part that turns redundancy from a spec-sheet number into something a customer feels, or rather does not feel. A facility without concurrent maintainability has to schedule an outage to service a UPS. A facility with it services the UPS on a Tuesday afternoon and nobody downstream is told, because there is nothing to tell them. The full requirement is set out in the Uptime Institute’s tier classification system, and we have covered what the Tier III standard actually commits a facility to in more detail.
There is a second half to the requirement that gets quietly skipped. Tier III does not only ask for redundant components. It asks for redundant, independent distribution paths. Four UPS units sitting on one shared distribution path are not truly redundant, because a fault in that path reaches all four. This is the distinction that separates a facility that merely owns backup equipment from one engineered not to go down.
Why are power redundancy and fibre redundancy the same idea?
The parallel is direct, not a metaphor. We have written before about why our network runs on five carriers instead of one, and why the shortest path is chosen dynamically rather than fixed. Electricity answers to identical logic, because a server does not care whether it lost connectivity or lost electricity. Both problems reduce to one question: what happens when a path fails, and does the answer involve anyone noticing?
| Fibre and network | Power |
| Five carriers, no single upstream dependency | N+1 UPS and generator units, no single source dependency |
| Diverse physical entry points, so one dig cannot sever every path | Independent distribution paths, so one fault cannot propagate to the other |
| Dynamic path selection across carriers for the fastest route | Automatic transfer switches move load with no human in the loop |
| A carrier outage never reaches your traffic | A UPS or generator failure never reaches your load |
In both cases the redundancy only counts if it is independent, not merely duplicated. Five carriers whose cables all enter the building through the same conduit are not redundant, which is why the physical route a cross-connect at 151 Front takes matters as much as the carrier on the end of it. Four UPS units on one distribution path are not redundant either. Independence is the whole point, and it is the expensive half.
How does power redundancy work at 151 Front Street West?
Two things are true here and they are worth separating, because most providers let them blur. 151 Front Street West is a Tier III facility. Amanah’s TOR1 and TOR2 suites sit inside it, and the power and cooling infrastructure serving those suites is built and operated to the same Tier III requirements.
That distinction matters. Being in a Tier III building and buying a Tier III deployment are not the same claim, and a provider who only makes the first one is describing their landlord, not their engineering. Ours runs N+1 across power and cooling on independent distribution paths, live and load-tested rather than filed under disaster recovery.
Cooling draws from Enwave deep lake water cooling, a purpose-built district system rather than a standard chiller plant, with redundant capacity of its own.
- N+1 redundancy across power, cooling and network, live and load-tested
- 15kW per rack in TOR1
- 30kW per rack in TOR2
Those 15kW and 30kW per-rack ceilings are the reason higher-density compute does not force a redesign of the power plan, which is the constraint behind our high-density colocation suites. The same design logic that keeps a UPS bank from being a single point of failure is what keeps our carrier mix from being one. Neither system asks you to trust an assumption. Both are load-tested, both are independent, and both are built so that maintenance and failure look identical to your equipment: invisible.
Walk the UPS bank, the generator plant and the distribution paths.
Every claim in this piece is something you can stand in front of. Book a tour of 151 Front Street West and see the redundant power and cooling paths behind it.
Questions about this topic
A UPS is battery-based and is always feeding your rack. If the utility drops, the UPS sustains that load on battery, typically for a couple of minutes, while the automatic transfer switch calls in the backup generator. The generator takes longer to start, and once running it sustains the panel, and through it the UPS, for one to two weeks on the fuel held on site and longer with resupply. Only the UPS ever touches your rack directly. The generator’s job is to keep the panel and the UPS fed, not to take over the load itself.
At 151 Front Street West, one to two weeks on the fuel held on site, and indefinitely beyond that for as long as resupply keeps arriving. A generator’s own runtime is bounded by fuel, not by the machine, which is why the meaningful question to ask any facility is not the tank size but how much runtime it holds on site, whether it has a standing fuel contract, and how many generator units it runs.
It means buying and maintaining more equipment than the load strictly requires. The cost is real, and it is what separates a facility built to the Tier III standard from one that is not. The return is that scheduled maintenance and unplanned component failure both become non-events for anyone connected to the system.
No. N+1 is one requirement inside Tier III, not the whole standard. Tier III also requires that the redundant components sit on independent distribution paths, not extra units bolted onto a single shared path. Redundancy without independent paths still leaves you exposed to one fault propagating through the path everything shares.
Most managed rack PDUs offer per-outlet or per-circuit metering, so you can see exactly how much power your equipment is drawing in near real time. That visibility is what lets you plan capacity, catch a runaway process before it trips a breaker, and right-size future deployments instead of guessing.
Nothing, if the facility is genuinely built to the Tier III standard. Concurrent maintainability means a UPS or generator unit can be taken offline for service while the N+1 units already in place absorb the load. There should be no maintenance window and no advance warning about a brief interruption.