Medium and small data centres are the backbone of enterprise IT, government agencies, healthcare facilities, and educational institutions worldwide. While they may not match the scale of hyperscale facilities, their power reliability requirements are equally demanding — and in many ways, more complex to engineer due to tighter budgets, limited space, and the constant pressure to scale.

Table of Contents

  1. What Is a Medium or Small Data Centre?

  2. Key Power Reliability Challenges

  3. Core Infrastructure Components

  4. Central vs. Distributed UPS Configuration

  5. Choosing the Right UPS

  6. The Modular UPS Advantage

  7. Precision Cooling Considerations

  8. Monitoring and Management

  9. Best Practices for Deployment

  10. Summary

1. What Is a Medium or Small Data Centre?

A medium or small data centre typically refers to a facility housing anywhere from a single server rack up to several dozen racks, serving a single organisation or a defined group of tenants. These are commonly found in corporate server rooms, university IT suites, hospital data hubs, regional government facilities, and colocation spaces for small and mid-sized businesses.

Unlike hyperscale data centres that operate tens of thousands of servers with multi-megawatt power feeds and fully redundant N+N infrastructure, medium and small data centres typically operate with a more constrained set of resources. Power budgets are measured in tens or hundreds of kilowatts rather than megawatts, and engineering teams are smaller — sometimes a single IT manager responsible for everything from hardware to power continuity.

Despite these differences, the criticality of the equipment they house is no less significant. A regional hospital's patient records system, a university's exam platform, or a company's ERP infrastructure must remain available around the clock. The consequences of unplanned downtime are real and measurable.

1–20

Racks

Typical small data centre

20–100

Racks

Typical medium data centre

10–500kW

IT Load

Common power envelope

2. Key Power Reliability Challenges

The structural characteristics of medium and small data centres create a distinct set of power challenges that must be addressed by the UPS and power distribution architecture.

Lower redundancy headroom

Large data centres routinely deploy 2N redundancy — full duplication of every power path. For smaller facilities, this level of investment is often not economically justified. As a result, the reliability of individual UPS units and power components carries more weight. A single point of failure in a non-redundant or N+1 architecture has a greater impact than in a fully redundant system.

Growth and scalability pressure

Medium and small data centres frequently grow in unpredictable ways. A new business unit, an acquired company's infrastructure, or a sudden increase in compute demand can quickly overwhelm a fixed-capacity UPS that was sized for current load only. The power system must be designed with expansion in mind from the very beginning.

Mixed load types

Unlike large facilities with homogeneous server deployments, smaller data centres often protect a wide variety of equipment simultaneously — production servers, storage arrays, networking equipment, telecommunications systems, and sometimes industrial control systems. Each may have different power quality requirements and runtime needs.

Limited facilities expertise

Smaller facilities rarely have dedicated facilities engineers on site. IT teams manage power infrastructure alongside their primary responsibilities. This makes simplicity of management, clarity of monitoring, and ease of maintenance especially important factors in UPS selection.

Key takeaway

In medium and small data centres, power reliability must be built into the UPS itself rather than compensated for by extreme infrastructure redundancy. Component quality, modularity, and manageability are not optional extras — they are the primary design criteria.

3. Core Infrastructure Components

A complete power system for a medium or small data centre is built from several integrated components. Understanding how these work together is essential to making correct sizing and configuration decisions.

Component 1

Automatic Transfer Switch (ATS)

The ATS manages the transition between utility power and generator backup. It monitors incoming supply quality and switches automatically when utility power falls outside acceptable parameters. In a dual-feed setup, the ATS can also select between two utility feeds. Transfer times of 20–100 ms are typical for mechanical ATS units; static ATS units can switch in under 10 ms.

Component 2

Central Power Distribution Unit (PDU)

The central PDU aggregates and distributes power from the ATS to all downstream equipment including UPS units, telecom DC power supplies, and rack-level PDUs. It provides circuit protection, metering, and in intelligent variants, remote per-outlet switching and load monitoring. Correct PDU sizing and breaker coordination is essential to prevent nuisance tripping.

Component 3

UPS System

The UPS provides continuous, conditioned power to critical loads during utility outages and power quality events. In a medium or small data centre, the UPS may be a single high-capacity unit (central configuration) or multiple distributed units protecting different load groups. Double-conversion (online) topology is the standard for server room applications.

Component 4

Battery Banks

Battery banks store the energy that the UPS delivers during an outage. They may be integrated within the UPS chassis or housed in separate battery cabinets connected by DC bus links. Runtime is directly proportional to battery capacity and inversely proportional to load. Battery health management is critical — degraded batteries are the most common cause of UPS failure during an actual outage.

Component 5

Telecom DC Power System

Telecommunications and networking equipment often requires -48V DC power rather than AC. A dedicated DC power system (rectifier and battery backed DC bus) feeds these loads with appropriate voltage and ensures compatibility with telecommunications-grade reliability standards such as ETSI EN 300 132.

Component 6

Monitoring System

A centralised monitoring platform integrates data from all power and environmental systems — UPS status, PDU load readings, battery health, generator run status, temperature, humidity, and access control. DCIM (Data Centre Infrastructure Management) software provides real-time dashboards, historical trending, and automated alerting.

4. Central vs. Distributed UPS Configuration

One of the most consequential architectural decisions in designing a power system for a medium or small data centre is whether to deploy a single centralised UPS or multiple distributed UPS units protecting different load groups. Both approaches have distinct advantages and trade-offs.

Central UPS configuration

In a central configuration, a single high-power UPS unit (or a parallel set of UPS units) feeds all critical loads through the facility's power distribution network. Battery banks are co-located with the UPS in a dedicated power room or dedicated zone within the data centre.

AspectCentral UPSDistributed UPSManagement complexityLow — single systemHigher — multiple unitsPower distribution flexibilityLimitedHigh — per-load groupBattery maintenanceCentralised, easierDistributed, more complexScalabilityHigh (modular UPS)High (add units)Failure impactBroader (all loads)Contained (one group)Best forStable, well-defined loads; growing environments with modular UPSMixed load types; facilities requiring per-group power conditioning

For central configurations, a modular UPS is strongly recommended. Modular systems allow capacity to be added incrementally as the data centre grows, without replacing the entire UPS or interrupting service. This directly addresses the scalability challenge inherent in medium and small data centres.

For distributed configurations, the choice of UPS type per load group should reflect the nature of that load. Fixed, stable loads such as core networking and storage are well served by tower-form UPS units. Growing or variable loads — additional compute capacity, expanding virtualisation clusters — benefit from modular UPS units that can be scaled without disruption.

A well-designed power architecture separates the question of capacity from the question of configuration — choose the topology that fits your operational model, then choose the capacity that fits your load.

5. Choosing the Right UPS for Your Data Centre

Selecting the appropriate UPS requires evaluating several interdependent factors. No single specification — capacity, topology, form factor, or brand — can be assessed in isolation.

Topology: always online double conversion

For any data centre application — regardless of size — online double-conversion topology is the appropriate choice. In a double-conversion UPS, all power delivered to connected equipment passes through the rectifier and inverter continuously. The equipment never sees raw utility power, providing complete isolation from all power anomalies including sags, spikes, harmonics, and frequency deviations. Transfer time in the event of an inverter fault is effectively zero due to the presence of a static bypass.

Capacity sizing with the 80% rule

Calculate your total IT load in watts, accounting for all servers, storage, networking, and ancillary equipment. Apply a power factor correction based on your equipment's PSU specifications, then size the UPS to a maximum of 80% of its rated capacity under normal operating conditions. This headroom accommodates inrush currents during equipment startup, future load additions, and battery capacity degradation over time.

Minimum UPS sizing formula

Min UPS VA = Total Load (W) ÷ (0.80 × UPS Power Factor)

Runtime requirements

Runtime should be driven by your continuity strategy. If a diesel generator is available on site, the UPS must provide sufficient runtime to bridge the generator's start-up and stabilisation period — typically 30–90 seconds — plus a safety margin. If no generator is available, the UPS must support a complete graceful shutdown of all operating systems and storage, which typically requires 5–20 minutes at full load.

Input voltage range and output waveform

Specify a wide input voltage window (±20–25% of nominal) to prevent unnecessary battery cycling during minor voltage fluctuations. Always specify pure sine wave output — modified sine wave is incompatible with the switching-mode power supplies used in all modern servers and storage equipment and causes premature PSU failure through harmonic heating.

6. The Modular UPS Advantage for Growing Environments

Modular UPS architecture has become the preferred approach for medium and small data centres that anticipate growth or require high availability without the cost of full infrastructure duplication.

In a modular UPS system, a common cabinet or frame houses a bus structure, power electronics, and battery management hardware. Individual power modules — typically ranging from 5 kW to 50 kW per module depending on the platform — are inserted into the frame to build up total capacity. Each module operates independently and shares the load with other modules in the frame.

Pay-as-you-grow CapEx

Purchase only the capacity you need today. Add modules as load grows, avoiding the capital cost of oversizing at the outset. This is particularly valuable in environments where growth rate is uncertain.

Built-in N+1 redundancy

By installing one module more than required by the current load, you achieve N+1 redundancy within a single cabinet. If any module fails, the remaining modules carry the full load without interruption and without requiring a separate standby UPS unit.

Hot-swap maintenance

Failed or end-of-life power modules can be removed and replaced while the UPS continues supplying power to connected equipment — provided sufficient spare capacity exists among the remaining modules. This eliminates the maintenance window problem that affects monolithic UPS designs.

Improved efficiency at partial load

A modular UPS can put unused modules into sleep mode when load is low, maintaining efficiency at partial load conditions. This is particularly valuable in new deployments where IT load builds up over time rather than being present at full capacity from day one.

Recommendation

For central UPS configurations in growing data centres, a modular online UPS in the 10–200 kVA range offers the optimal balance of reliability, scalability, and total cost of ownership. For distributed configurations, combine modular UPS units for growing load groups with fixed tower UPS units for stable, well-defined loads such as core network infrastructure.

7. Precision Cooling Considerations

Power and cooling are inseparable in data centre design. The UPS efficiency directly affects the thermal load your cooling infrastructure must handle. A UPS operating at 92% efficiency at 50 kW load dissipates approximately 4 kW as heat into the room — a significant addition to the cooling budget that must be planned for.

Precision cooling systems for medium and small data centres are available in several formats, each suited to different space constraints and heat density profiles:

  1. 1

    Room-level precision air conditioners

    Cool the entire room volume. Suitable for lower-density deployments and smaller facilities. Easier to deploy but less efficient as rack densities increase.

  2. 2

    In-row precision cooling

    Cooling units deployed between rack columns, delivering cold air directly to rack inlets and capturing hot exhaust before it mixes with room air. More efficient at medium densities (5–20 kW per rack).

  3. 3

    In-rack cooling

    Rear-door heat exchangers or in-rack cooling distribution units handle heat at the rack level, ideal for high-density computing (20 kW+ per rack) where room-level or row-level cooling cannot keep pace.

The UPS should be positioned in a dedicated power area separated from the main server floor where possible, reducing the heat contribution to the primary computing zone and simplifying maintenance access. Battery banks are particularly sensitive to temperature — every 10°C above 25°C approximately halves the service life of VRLA batteries and significantly accelerates degradation in Li-Ion chemistries as well.

8. Monitoring and Management

Effective monitoring transforms a passive power infrastructure into an actively managed system. In medium and small data centres where staffing levels are lean, good monitoring software significantly reduces the risk of undetected failures developing into service-affecting events.

UPS network management

Every enterprise-grade UPS should be equipped with a Network Management Card (NMC) providing SNMP v3 access for integration with existing network monitoring platforms (Nagios, Zabbix, PRTG, SolarWinds, and similar). The NMC should also support secure HTTPS access for web-based management and provide configurable SNMP traps for all alarm conditions.

Graceful shutdown software

Connect the UPS to all protected servers via the network management card and appropriate shutdown agents. When a power event triggers the UPS to operate on battery, the shutdown software monitors remaining runtime and initiates an orderly shutdown of all virtual machines and operating systems before battery depletion. In VMware and Hyper-V environments, this includes live migration of workloads to unaffected hosts before local shutdown.

Environmental monitoring

Integrate temperature and humidity sensors, water leak detection strips, smoke detectors, and access control event feeds into the same monitoring platform as the UPS and PDU data. Comprehensive environmental awareness is essential in smaller facilities where a single cooling failure or water ingress event can affect the entire data centre simultaneously.

Monitor: Battery state of health

Track battery capacity, internal resistance, and temperature trends. Set alerts for batteries approaching end of life before they fail.

Monitor: Load percentage

Track load as a percentage of rated capacity over time. Rising trends indicate the need for capacity expansion before the 80% threshold is breached.

Monitor: Input voltage quality

Log the frequency and severity of input voltage anomalies to assess utility feed quality and identify when infrastructure investment in power conditioning is warranted.

Monitor: UPS internal temperature

Elevated internal temperatures indicate battery degradation, airflow obstruction, or overload conditions — all of which require prompt investigation.

9. Best Practices for Deployment

Applying established best practices during planning, installation, and ongoing operation significantly reduces the risk of power-related downtime in medium and small data centres.

10. Summary

A robust power system for a medium or small data centre is built on the integration of several key components: an automatic transfer switch managing utility and generator sources, a central PDU distributing power to all downstream equipment, a properly sized UPS providing continuous conditioned power, adequate battery capacity matched to runtime requirements, and a monitoring platform that makes the health of the entire system visible in real time.

The architectural choice between central and distributed UPS configuration should reflect the operational model of the facility and the nature of the loads being protected. For most growing environments, a central modular UPS represents the optimal combination of reliability, scalability, and manageability.

The modular UPS approach — deploying a scalable platform that grows with the data centre rather than being replaced by it — is particularly well suited to the realities of medium and small data centres, where growth is often unpredictable, budgets are constrained, and the operational overhead of managing complex infrastructure must be kept to a minimum.

Key principles recap