When a major Irish financial institution's data centre lost mains power at 2:47 AM on a Tuesday morning, the event that followed was not determined by how good the generator was or how fast the UPS transferred. It was determined by a design decision made two years earlier: whether to install one UPS or two. Redundancy is the architectural commitment that decides whether a single point of failure becomes an outage or an invisible, logged event. Understanding redundancy — what N, N+1 and 2N actually mean in practice — is the most fundamental design decision in any Irish data centre project.
Why Redundancy Defines Data Centre Availability
Availability is expressed as the percentage of time a system is operational. The difference between 99.9% (Tier I) and 99.999% (Tier IV) sounds small — but translates to the difference between 8.7 hours of unplanned downtime per year and 5.3 minutes. For an Irish cloud service provider with SLA penalties of €50,000 per hour of unavailability, this difference represents €420,000 per year in potential penalty exposure.
Availability is determined primarily by the redundancy of the critical path — power, cooling and network. Every component in the critical path has a Mean Time Between Failures (MTBF) and a Mean Time To Repair (MTTR). Redundancy eliminates single points of failure, so that the failure of any single component does not interrupt service while repair is underway.
Redundancy Terminology Defined
| Model | Description | Failures Tolerated | Tier Alignment | Capex Impact |
|---|---|---|---|---|
| N | Exactly the capacity required; no redundancy | 0 | Tier I | Baseline (1×) |
| N+1 | One spare component above the minimum required | 1 | Tier II–III | +20–40% |
| N+2 | Two spare components above the minimum required | 2 | Tier III+ | +35–55% |
| 2N | Fully duplicated independent systems (2 × N capacity) | N (entire primary system) | Tier IV | +80–100% |
| 2(N+1) | Two independent N+1 systems; ultimate redundancy | N+1 entire primary system | Tier IV (extreme) | +150–200% |
Power Path Redundancy: From Utility to Rack
The power path in a data centre runs from the EirGrid utility connection through multiple transformation and conditioning stages to individual server power supplies. Redundancy must be designed at every level to eliminate single points of failure on the critical path.
Utility Supply Redundancy
Large Irish data centres (above 5 MW) typically negotiate dual utility connections from EirGrid — supplied from separate substations on independent cable routes. EirGrid's Grid Connection Policy requires applicants to demonstrate that dual feeds come from independent portions of the transmission network, not just separate circuits from the same substation. At Cork's Tivoli industrial estate, for example, operators with dual connections route from the Cork South 110 kV substation and the Carrigaline 38 kV substation respectively to achieve genuine independence.
UPS Redundancy
UPS systems are configured in three primary redundancy architectures:
- Standalone (N) — a single UPS serves the entire load. Failure causes immediate outage. Used only in Tier I or low-criticality environments.
- Parallel redundant (N+1) — multiple UPS modules operate in parallel; one module can fail without load impact. Example: four 500 kW UPS modules serving a 1.5 MW load (N=3, one spare module)
- Fully redundant (2N) — two completely independent UPS systems, each sized to carry 100% of the load. Servers have dual power supplies: PSU A connects to UPS System A; PSU B connects to UPS System B. Either complete UPS system can fail without impact.
Generator Redundancy
Diesel generators provide extended backup power beyond the UPS battery runtime. Generator redundancy configurations parallel those of UPS systems, with additional considerations specific to rotating machinery:
- Single generator (N) — adequate for small facilities; any failure causes outage after battery depletion
- N+1 generators — the most common configuration for Irish Tier III colocation; one generator can fail or be taken offline for maintenance without load impact
- 2N generators — two independent generator sets, each rated for 100% of load; required for Tier IV; each set has its own fuel tank, ATS (Automatic Transfer Switch) and starting system
For N+1 and 2N generator sets, synchronisation is critical. Generators must be synchronised (matched in voltage, frequency and phase angle) before paralleling onto the common LV busbar. Modern governor and AVR controllers achieve synchronisation within 5–10 seconds of generator run-up — critical for maintaining supply continuity when the utility supply fails and UPS batteries are depleting.
Irish planning requirements (An Bord Pleanála and local authority planning conditions) typically limit above-ground diesel storage to 2,500 litres per tank before fire engineering justification is required. For a 2 MW facility with N+1 generators consuming approximately 500 litres/hour at full load, the standard design provides 48 hours of fuel autonomy through multiple storage tanks with secondary containment.
Cooling Redundancy
Cooling failure causes rapid temperature rise in a data hall. At a typical server inlet temperature of 25 °C, a complete cooling failure in a 2 MW data hall will cause rack inlet temperatures to exceed the ASHRAE A2 envelope (35 °C) within 3–8 minutes depending on room thermal mass. This is below the typical generator start-and-transfer time, meaning cooling redundancy cannot rely on the generator as a backup — it must be inherent in the cooling plant design.
CRAC/CRAH Unit Redundancy
Data hall cooling units (CRACs or CRAHs) are typically deployed in N+1 or 2N configurations. For N+1, units are sized and positioned so that if any single unit fails, the remaining units increase fan speed to maintain cooling — a function managed automatically by the DCIM or BMS. For 2N, two independent cooling systems (separate chiller plants, separate pipework rings, separate power feeds) serve the same data hall.
Irish Climate Consideration: Free Cooling
Ireland's mild maritime climate — Dublin's average annual temperature is 9.8 °C — provides significant free cooling hours. Irish data centres using adiabatic coolers or direct outside air cooling (DOAC) systems can achieve 6,000–7,000 hours per year of full or partial free cooling, dramatically reducing mechanical chiller run hours and compressor maintenance requirements. Redundancy design for free-cooling systems must consider the failure mode of the adiabatic media — typically evaporative pads — and ensure that mechanical cooling can assume 100% of the load if the adiabatic system is offline.
Network Redundancy
Network redundancy is often the overlooked dimension of data centre resilience design. A 2N power and cooling plant provides no benefit if a single switch failure takes down all connectivity. Network redundancy is designed at three levels:
- Top-of-Rack switching — dual ToR switches per rack row with LACP (Link Aggregation Control Protocol) bonding; server NICs connect one port to each ToR switch
- Core switching — dual core switches with OSPF equal-cost multi-path (ECMP) routing; both paths active simultaneously
- Internet connectivity — dual upstream providers, BGP multi-homed; if one ISP fails, traffic fails over to the second provider within BGP convergence time (typically 30–60 seconds)
For Irish facilities, internet connectivity diversity is significant. Dublin's primary internet exchange — the INEX (Internet Neutral Exchange) — is located in two facilities with full mesh connectivity. Operators connected to INEX at both locations achieve effective dual-path connectivity to most European networks without dependency on a single ISP.
Concurrent Maintainability and Fault Tolerance
Concurrent maintainability (Uptime Institute Tier III requirement) means that any component in the critical path can be taken offline for planned maintenance without affecting IT load. This requires N+1 or better on all critical components — because maintenance of one component reduces the system to N capacity, and any simultaneous failure would cause an outage. Tier III facilities typically achieve 99.982% availability.
Fault tolerance (Tier IV requirement) goes further: the facility must sustain any single fault — planned or unplanned — without impact to IT load. This requires 2N on all critical paths, because any component failure leaves a complete N system still operational. Tier IV facilities achieve 99.995% availability.
The Cost of Redundancy
Every level of redundancy above N has a direct capital cost. The relationship is not always linear. Moving from N to N+1 adds approximately 20–30% to the cost of the affected system. Moving from N+1 to 2N typically adds 60–80% above N+1 — because 2N requires not just additional components but entirely separate infrastructure: separate switchrooms, separate cable risers, separate UPS rooms and separate cooling plant rooms to ensure physical independence. A single fault on a 2N system serving 100 racks should be solvable by inspection of a single room — not by chasing shared infrastructure.
For a 5 MW Irish data centre, the additional capital cost of moving from N+1 to 2N across power, cooling and network infrastructure is typically €8M–€15M. Justified against the cost of a single major outage — which for a Tier III colocation facility serving financial services customers can reach €2M–€5M per event in SLA penalties and customer churn — the payback period is measured in avoided incidents rather than years.
Frequently Asked Questions
N+1 redundancy means that for every N components required to support the full load, one additional (spare) component is provided. If any single component fails, the remaining N components continue to carry the full load without interruption. For example, if four UPS modules (N=4) are required to support the IT load, an N+1 system installs five modules — one failure can be absorbed without any loss of capacity.
2N redundancy approximately doubles the capital cost of the power and cooling infrastructure but is justified when the cost of downtime exceeds the capital cost premium over the facility lifetime. For Irish hyperscale and colocation operators with SLAs guaranteeing 99.999% availability (Tier IV equivalent), 2N is typically contractually required. For enterprise data centres with less stringent uptime requirements, N+1 often offers the best balance of resilience and capital efficiency.
Dublin's major colocation operators — including Equinix, Digital Realty and CyrusOne — advertise Tier III or Tier IV equivalent designs. Tier III requires N+1 across all critical systems with concurrent maintainability. Tier IV requires 2N or higher across all critical paths with full fault tolerance. Most operators offer both tiers across different product suites, with Tier IV commanding a premium of 20–35% on colocation pricing.
Need Data Centre Redundancy Design for an Irish Project?
ASDV's engineers design N+1 and 2N power, cooling and network architectures for Irish data centres — from concept redundancy strategy through to detailed single-line diagrams and schematic design packages.