Technical Note: Data Centres

Modern data centres demand integrated engineering solutions that prioritise resilience, operational continuity and long-term performance. This technical note explores the critical design considerations that shape mission-critical facilities, including electrical infrastructure, mechanical cooling, redundancy, sustainability and scalability. Whether delivering hyperscale, enterprise or edge facilities, understanding these engineering principles is essential for creating reliable digital infrastructure.

Engineering Considerations for Data Centre Design

Data centres represent one of the most technically demanding sectors within the built environment. Unlike conventional commercial facilities, these mission-critical assets are engineered to support continuous digital infrastructure with availability requirements typically exceeding 99.999%. Consequently, early design and construction decisions directly dictate operational resilience, energy efficiency, maintainability, and lifecycle costs.

The rapid expansion of artificial intelligence (AI), cloud architecture, edge computing, and digital transformation has dramatically escalated demand for high-performance infrastructure. Modern facilities must now accommodate unprecedented rack power densities, higher heat loads, and sophisticated control systems, all while adhering to tightening sustainability frameworks and regulatory requirements.

Successful data centre projects therefore rely on more than simply specifying redundant equipment. Long-term performance is determined by how effectively electrical distribution, mechanical cooling, communications infrastructure, building controls, and operational procedures are integrated into a cohesive solution capable of supporting continuous operation throughout the asset lifecycle.

Ultimately, the overarching engineering design must align precisely with the facility’s business risk profile. This requires balancing core availability, redundancy, and resilience targets alongside design of appropriate power generation, thermal management, physical security, and fire protection.  All while ensuring the facility remains scalable and sustainable in operation.

Data Centre Requirements and Tier Classifications

Data centre infrastructure operates under a fundamental imperative: ensuring the continuous, uninterrupted processing and storage of digital information. To navigate the complex balance between capital expenditure, risk mitigation, and continuous operation, engineering decisions are evaluated through three distinct but interconnected concepts:

• Redundancy: the physical provision of backup components and distribution pathways (the static hardware strategy).
• Resilience: the architectural and operational ability to absorb, adapt to, and recover from real-time faults, transients, and external environmental shocks without dropping critical IT load i.e. a single point of failure.
• Availability: the ultimate business metric, expressed as a percentage of operational uptime, that measures the time a service is actually usable and performing as intended.

The practical application of these principles is formally codified through standardised design frameworks, most notably the Uptime Institute Tier Classification System.

Uptime Institute Tier Classification System

The Uptime Institute Tier Classification System was established in 1995, as a universal benchmark, to create a consistent global standard for measuring data centre reliability and performance.  The tier system translates complex engineering into a standardised language which allows businesses, investors, and service providers to instantly understand a facility’s reliability and risk profile to justify pricing and structure the Service Level Agreements (SLAs).

For engineering and operations teams, the framework acts as a benchmark for electrical and mechanical redundancy needs, ensuring that a single pipe leak or power failure won’t take down the entire network. Once the building is live, site managers use Tier principles to plan safe maintenance schedules and train staff, ensuring routine upgrades can happen without disrupting the live digital services that modern businesses depend on.

The framework categorises data centres into four distinct Tiers based on their underlying mechanical and electrical architecture, concurrent maintainability, and fault tolerance as follows:

Summary Comparison of Uptime Institute Tiers

Metric / Feature	Tier I	Tier II	Tier III	Tier IV
	Basic	Redundant Components	Concurrently Maintainable	Fault Tolerant
Primary Focus	Cost-effective, basic capacity	Line-of-site redundancy	Active maintenance with zero downtime	Maximum resilience against any failure
Uptime Guarantee	99.671%	99.741%	99.982%	99.995%
Max Annual Downtime	~28.8 hours	~22.7 hours	~1.6 hours	Under 26 minutes
Distribution Paths	Single path	Single path	Multiple paths (1 active, 1 passive)	Multiple paths (All active simultaneously)
Redundancy Level	N	N + 1 (Component redundancy)	N + 1 (Component & path redundancy)	2(N + 1) or S + S (Full isolation)
Impact of Maintenance	Complete system shutdown required	Complete system shutdown required	Zero impact on live IT operations	Zero impact on live IT operations
Impact of Unplanned Error	Complete system crash	Complete system crash	Potential disruption if critical path fails	Zero impact (System absorbs the error)
Best Used For	Small businesses, startups, test labs	Regional offices, non-critical backups	E-commerce, cloud providers, healthcare	Global banks, defense, vital infrastructure

Resilience vs Redundancy

Infrastructure resilience is the capacity of a data centre to sustain critical operations across all activities during equipment failures, maintenance activities, utility interruptions and other foreseeable disruptive events including electrical, mechanical, ICT, fire protection and control systems.

Resilience must not be confused with simple redundancy; while redundancy merely duplicates components or paths to add capacity, resilience is the facility’s systemic ability to maintain operations under degraded conditions. Achieving true resilience requires a unified lifecycle approach from concept design through to commissioning, demanding close interdisciplinary coordination to ensure that engineering decisions in one discipline do not inadvertently compromise another.

Single Points of Failure

A fundamental objective of resilient architecture is the systematic elimination of Single Points of Failure (SPOFs), where the loss of a single component, connection, or system drops the critical load. Beyond major equipment, SPOFs frequently exist within shared infrastructure, such as:

• Common electrical switchboards.
• Shared fuel systems.
• Common control networks.
• Single communications pathways.
• Shared chilled water headers.
• Single Building Management System (BMS) servers.
• Common fire alarm interfaces.
• Network switches or control processors.

Mitigating these risks requires structural diversity rather than mere duplication. Designers must establish physical and logical separation, such as independent A/B distribution paths, separated cable routes, diverse telecom entries, and separate cooling circuits, to ensure redundant systems do not rely on common utilities or share common pathways that could be compromised simultaneously by a single physical event.

Fault Containment

Fault containment is equally important to fault prevention.  Electrical protection systems should be coordinated to ensure faults are isolated as close as practicable to the point of failure, preventing unnecessary upstream tripping and loss of supply to unaffected loads. Protection coordination studies, discrimination analysis and arc flash assessments should therefore form part of the electrical design process.

Similarly, mechanical systems should incorporate sectional isolation valves, independent pipework branches and control strategies capable of maintaining partial operation following component failure. The objective is not simply to prevent failures, but to limit the consequences when failures inevitably occur.

Operational Resilience

Physical infrastructure alone cannot guarantee continuous uptime; true resilience extends heavily into the operational domain. Site operators must thoroughly understand and anticipate how the entire facility behaves under dynamic stress events, including utility blackouts, generator start-ups, UPS battery discharges, maintenance bypass transitions, and automatic transfer sequences. Because functional dependencies often remain hidden during isolated component testing, these complex operational scenarios must be rigorously validated through full-scale, integrated systems testing during the commissioning phase to guarantee the facility performs reliably under real-world emergency conditions.

Redundancy Strategy

Redundancy provides additional infrastructure capacity or alternative service paths that enable critical operations to continue during equipment failures or planned maintenance.  Appropriate redundancy improves operational resilience while supporting maintainability; however, excessive redundancy may introduce unnecessary complexity, increased lifecycle costs and additional operational risks.  Redundancy should therefore be applied strategically based on engineering assessment rather than uniformly across all building services.  The most common redundancy configurations include:

Configuration	Description
N	No redundant capacity
N+1	One additional redundant component
N+2	Two additional redundant components
2N	Two completely independent systems
2(N+1)	Two independent systems, each with additional redundant capacity

Each configuration offers different levels of resilience, maintainability and fault tolerance as per the tier classification system.

Electrical and Mechanical Redundancy Integration

Electrical redundancy is integrated into the power systems via dual utility feeds, parallel transformers, standalone uninterruptible power supply (UPS) systems, standby generators, and static transfer switches (STS) feeding dual-corded servers. To neutralise common-mode failure risks, strict physical and logical separation must be enforced between redundant power paths wherever practicable.

Mechanically, redundancy is established via N+1 chillers, duty/standby pump sets, redundant cooling towers, additional duty standby computer room air handler (CRAH) units, and independent chilled/condenser water circuits. The specific engineering topology selected for the mechanical plant must directly account for localised equipment reliability metrics, Mean Time to Repair (MTTR), maintenance cycles, and the unique thermal decay characteristics of the active data halls.

Resilience and Redundancy Design Trade-Offs

Maximising system resilience and redundancy correlates with trade-offs across the facility’s lifecycle. Escalating redundancy configurations directly increases capital expenditure (CapEx), ongoing operational energy consumption (OpEx), plantroom spatial requirements, and the volume of preventive maintenance activities.

Importantly, it increases the asset’s embodied carbon footprint and intensifies control system complexity, which can elevate the risk of human error or automated sequence failures. Qualified engineering judgment must therefore carefully weigh incremental availability gains against the holistic, long-term operational performance and efficiency of the facility.

Engineering System Selection

Selecting infrastructure topologies requires balancing reliability, efficiency, scalability, and lifecycle cost, with final selection needing to adapt to localised constraints, rack densities, and specific business risk profiles.  Designers will review different options through the concept design stage of a project, before progressing into detailed design of the selected systems / plant and equipment.

System / equipment selection influences available floor space, structural loading, maintenance requirements, lifecycle cost, replacement frequency, thermal management and operational resilience. Selection should consider design life, discharge characteristics, charging performance, ambient operating conditions and manufacturer support.

Electrical Infrastructure Topologies

• Uninterruptible Power Supply (UPS) – Lithium-Ion vs. Valve-Regulated Lead-Acid (VRLA)
  • Lithium-Ion: Offers high energy density, a small footprint, a long lifespan (10–15 years), and reduced maintenance. However, it incurs high initial capital expenditure (CapEx) and requires complex Battery Management Systems (BMS) with strict thermal runaway protections.
  • VRLA: Features a low initial purchase cost, mature and deeply understood technology, and simple recycling paths. On the downside, it requires frequent testing, demands heavy floor reinforcement, carries a short service life (3–5 years), and exhibits high sensitivity to ambient temperature spikes.
• Standby Generation – Diesel vs. Gas vs. Hydrotreated Vegetable Oil (HVO)
  • Diesel: Provides rapid step-load acceptance, high energy density, and mature global supply chains. Conversely, it produces high localised emissions, demands strict environmental compliance, and risks fuel contamination or microbial growth over long storage periods.
  • Gas (Turbine/Engine): Delivers clean emissions, removes on-site fuel storage risks, and operates quietly. However, it suffers from poor step-load response, relies heavily on vulnerable public utility gas networks, and demands long start-up sequences.
  • HVO (Bio-Diesel): Acts as a drop-in replacement for standard diesel that slashes net carbon emissions by up to 90% without requiring infrastructure modifications. Its main drawbacks are constrained global supply chains, market price volatility, and potential long-term shelf-life degradation.
• Power Distribution – Busbar Trunking vs. Traditional Cable and Conduit
  • Busbar Trunking: Highly scalable with plug-and-play tap-off boxes, space-efficient, and easy to modify for future rack expansions. The disadvantages include higher initial component costs and a susceptibility to widespread dust or water contamination if not properly IP-rated.
  • Cable & Conduit: Offers low initial material costs, utilises standard electrical trade skills, and provides excellent physical routing flexibility. However, it congests under-floor or overhead spaces, restricts airflow, and requires labor-intensive modifications when scaling up rack densities.

Mechanical Cooling Topologies

• Chilled Water (ChW) Systems (Centralised Air/Water-Cooled Chillers & CRAHs)
  • Pros: Highly efficient at scale, easily supports variable-speed primary pumping, accommodates high-density loads, and integrates well with waterside economisers for free cooling.
  • Cons: High architectural complexity, risks water leaks near IT equipment, requires large plant-rooms, and demands complex failure recovery sequences during power transitions.
• Direct Expansion (DX) Systems (Perimeter CRACs)
  • Pros: Low capital cost, simple independent operation per unit, eliminates fluid loops near IT racks, and requires minimal engineering oversight.
  • Cons: Poor energy efficiency at partial loads, limited capacity scaling, and struggles to support modern High Performance Computing (HPC) or Artificial Intelligence (AI) rack densities.
• Direct-to-Chip Liquid Cooling & Immersion Cooling
  • Pros: Capitalises on water’s high volumetric heat capacity to handle extreme densities (50–150+ kW per rack), slashes fan energy consumption, and enables high chiller-water return temperatures for maximised free cooling.
  • Cons: Extreme initial capital expenditure, introduces fluids directly into the IT chassis, requires specialised server hardware, complicates standard maintenance access, and demands deep interdisciplinary design coordination.

Fire Suppression Topologies

• Gaseous / Clean Agent Systems (e.g., Novec 1230, Inergen, FM-200)
  • Pros: Extinguishes fires rapidly by disrupting the chemical reaction or reducing oxygen levels without causing any structural or electrical damage to active IT hardware.
  • Cons: Requires strict room integrity and enclosure pressure testing to hold the gas concentration, carries high recharge costs after discharge, and can generate destructive acoustic levels that damage hard disk drives (HDDs) if specialised silencing nozzles are omitted.
• Pre-Action Water Sprinkler Systems (Single or Double Interlock)
  • Pros: Highly reliable for structural fire protection, cost-effective, and safe for building occupants. The interlocking mechanism ensures pipes remain dry until both smoke detection and thermal fuse links activate, mitigating accidental discharge risks.
  • Cons: Any water discharge permanently destroys active, unsealed electronic components, cleanup operations cause extensive business disruption, and the mechanical pipe networks demand intensive localised spacing coordination with overhead cable trays and ductwork.

Maintainability

Maintainability is the ability to inspect, test, service, repair or replace infrastructure safely and efficiently throughout its operational life. Within mission-critical facilities, maintainability extends beyond equipment accessibility to encompass the ability to perform planned maintenance without compromising service availability.

Concurrent Maintainability

Concurrent maintainability enables any individual component to be removed from service while maintaining full support for the critical load. This influences all aspects of engineering design including electrical power feeds and switchboard configurations, pipework arrangements and valve locations, maintenance bypasses / equipment isolation and controls.  Where concurrent maintainability is required, maintenance activities should not require complete shutdown of critical systems.

Documentation for Maintainability

Regardless of how well a facility has been designed, maintenance activities cannot be undertaken safely or efficiently without reliable documentation describing the installed infrastructure, its operational intent and its interdependencies.

For mission-critical facilities such as data centres, maintenance personnel must be able to rapidly identify equipment, understand system topology, verify isolation points and assess the operational consequences of taking assets out of service. Inaccurate or incomplete documentation increases maintenance duration, introduces unnecessary operational risk and may result in incorrect isolation, equipment damage or unplanned service interruption.

A comprehensive lifecycle documentation package should typically include:

• As-built drawings showing the final installed condition of all building services, including electrical single-line diagrams, mechanical schematics, pipework layouts, communications pathways and equipment locations.
• Operation and Maintenance (O&M) Manuals providing manufacturer recommendations, system descriptions, operating procedures, maintenance requirements, fault-finding guidance and warranty information.
• Building Handover Manuals, bringing together project-wide engineering information into a structured reference that supports the transition from construction to operation.
• Asset registers containing unique equipment identifiers, manufacturer details, model numbers, serial numbers, design capacities, installation dates, maintenance intervals and replacement information.  

The value of these documents extends well beyond project handover. As infrastructure is modified, expanded or refurbished throughout its operational life, engineering documentation becomes the primary reference for designers, contractors, commissioning engineers and maintenance personnel. Maintaining documentation that accurately reflects the installed condition of the facility reduces operational risk, shortens maintenance outages and supports informed engineering decision-making.

For facilities designed with concurrent maintainability or fault-tolerant infrastructure, documentation assumes an even greater level of importance. Maintenance teams must understand not only the location and function of individual assets, but also the relationship between redundant systems, alternate power paths, cooling circuits and control strategies. Without accurate engineering information, the resilience designed into the facility can be unintentionally compromised during routine maintenance activities.

Scalability and Future Expansion

Data centres should be designed to accommodate future increases in IT demand with minimal disruption to existing operations. Scalability enables owners to respond to changing technologies, increased computing density and evolving business requirements while protecting long-term capital investment. Unlike conventional commercial developments, data centres frequently undergo multiple infrastructure expansions throughout their operational life.

Modular Designs

Modular infrastructure enables capacity to be installed progressively such as modular UPS systems, prefabricated plant modules, containerised power systems, modular chillers and busway electrical distribution.  Modular deployment improves capital efficiency while reducing commissioning complexity and construction risk.

Designing for AI

Artificial Intelligence is fundamentally changing data centre infrastructure planning by significantly increasing rack power densities. These higher-density computing environments require greater electrical capacity, more advanced cooling solutions and supporting infrastructure capable of accommodating increased thermal loads while maintaining resilience, maintainability and energy efficiency.  Providing sufficient flexibility during initial design significantly reduces future upgrade costs.

Advanced Sustainability, Regulatory Compliance, and Grid Interaction

Sustainability has rapidly become one of the defining challenges in modern data centre design. The unprecedented growth of cloud computing, Artificial Intelligence (AI), High-Performance Computing (HPC) and digital services has accelerated demand for data centre capacity worldwide, placing increasing pressure on electricity networks, water resources and carbon reduction commitments.  This rapid expansion of hyperscale developments has prompted greater scrutiny from governments, regulators, electricity providers and local communities and increasingly influencing planning approvals, utility connection agreements and investment decisions.

Navigating Global Regulatory and Connection Bans

Unchecked resource consumption has triggered aggressive regulatory and planning interventions globally, making strict environmental compliance a core prerequisite for securing utility connections:

• Construction Moratoriums: Major digital hubs like New York State have instituted strict freezes on hyperscale data centre construction drawing 50 MW or more to protect grid stability. Similar long-term development bans exist across European metros like Amsterdam.
• On-Site Generation Mandates: Grid operators, such as Ireland’s EirGrid, have heavily restricted grid connections, forcing new facilities to provide independent, on-site power generation to avoid draining the civilian network.
• National AI and Utility Legislation: Regulatory bodies, including the Clean Energy Council and national governments, are enforcing new grid connection standards and establishing dedicated offices to regulate exactly where data centres can be built, how much water they can consume, and how they must support the clean energy transition.

Sustainability Beyond Operational Energy

Historically, sustainability initiatives focused primarily on reducing operational electricity consumption of a development. However, there is now a broader shift from measuring operational efficiency alone towards evaluating the total environmental impact of data centre infrastructure across its entire lifecycle, wider sustainable initiatives must be considered:

• Embodied Carbon: Reducing emissions associated with construction materials, structural systems, electrical equipment, batteries and mechanical plant.
• Circular Economy Principles: Designing infrastructure to facilitate refurbishment, modular upgrades, component reuse and responsible end-of-life disposal.
• Demand Response and Grid Support: Data centres should transition from passive energy consumers into active grid assets. Utilising large-scale solar generation paired with Battery Energy Storage Systems (BESS) allows facilities to participate in demand response programs, executing peak-shaving during high public demand or injecting power back into the network to act as a shock absorber for local grids.
• Minimising Water Consumption: Traditional evaporative cooling methods place an unsustainable burden on local drinking water supplies. Modern facilities are eliminating water consumption entirely by deploying high-efficiency closed-loop direct liquid cooling, airside economisers, or direct-to-chip technologies that preserve vital local water security.
• Waste Heat Recovery: Capturing rejected heat from data centre cooling systems for district heating, commercial developments or industrial processes where viable.
• Refrigerant Selection: Selecting lower Global Warming Potential (GWP) refrigerants that comply with evolving environmental legislation while maintaining system efficiency and reliability.
• Decarbonisation: Standby generation is abandoning fossil diesel in favour of sustainable Hydro-treated Vegetable Oil (HVO). Major operators, including Amazon Web Services (AWS) have transitioned, or are actively transitioning, their backup systems away from traditional fossil diesel to 100% HVO (HVO100).

Conclusion

The design and delivery of modern data centres requires a holistic engineering approach that extends far beyond the selection of individual mechanical and electrical systems. As digital infrastructure continues to evolve through artificial intelligence, cloud computing, and high-performance computing, facilities must be designed to accommodate increasing power densities, complex cooling requirements, evolving operational demands, and increasingly stringent sustainability expectations.

Achieving reliable and resilient operation requires careful consideration of redundancy, fault tolerance, maintainability, scalability, and lifecycle performance from the earliest stages of design. Engineering solutions must be aligned with the operational risk profile of the facility, ensuring that infrastructure provides the required level of availability without introducing unnecessary complexity, cost, or environmental impact.

The elimination of single points of failure, effective fault containment, concurrent maintainability, and rigorous integrated systems testing are fundamental to ensuring that data centres can continue operating during planned maintenance activities and unexpected failures. Equally important is the provision of accurate lifecycle documentation, including as-built information, operation and maintenance manuals, asset registers, and system records, which enables operators to safely maintain, modify, and expand critical infrastructure throughout its operational life.

Future data centres must also be designed with flexibility in mind. The rapid growth of AI and high-density computing means that infrastructure installed today must support future technology requirements without significant disruption or premature replacement. Modular design strategies, adaptable power and cooling systems, and consideration of future expansion requirements are therefore essential components of sustainable long-term planning.

Ultimately, a successful data centre is not defined solely by its Tier classification or installed redundancy levels, but by how effectively its engineering systems, operational processes, documentation, and maintenance strategies work together as a complete lifecycle solution. By integrating resilience, maintainability, scalability, and sustainability into every stage of development, data centre owners and operators can achieve secure, efficient, and future-ready facilities capable of supporting the digital infrastructure demands of tomorrow.