Data Centre Cooling & UPS Power Australia
Industry Solutions / Data Centres and IT Infrastructure
Data Centre Cooling and UPS Power for Australian Data Halls
Data centre cooling is the removal of IT heat from a data hall using CRAC or CRAH units, chilled water or condenser water plant and aisle containment, and in Australia it typically accounts for 30 to 40 per cent of total facility energy at a whole-of-facility PUE between 1.3 and 1.6. Indigi Power and Cooling designs, installs, commissions and maintains the room-level cooling and uninterruptible power that colocation providers, enterprise data halls and government facilities run on, from a single 20 kW comms room through to multi-megawatt N+1 and 2N sites.
A data centre is not a room with air conditioning in it. It is a thermal and electrical system with a defined redundancy topology, a maintenance regime that has to be executable without taking load offline, and a set of measured numbers that the business is held to. Most of the failures we are called to were designed in years earlier: a cooling plant that is nominally N+1 but shares a single chilled water header, a UPS that cannot be bypassed without a shutdown, or a room whose CRAC units have been quietly running at 100 per cent since the last three racks went in.
The Australian picture has its own shape. The Department of Climate Change, Energy, the Environment and Water notes that the average Australian data centre is now over 20 years old, and a large share of the national footprint is not hyperscale at all. It is enterprise data halls, regional government facilities, university computing rooms and colocation suites built in the 2000s and progressively loaded well past their design intent. Those sites are where the real work is, and they are the sites where redundancy claims and reality most often diverge.
Indigi Power and Cooling is an Indigenous and Veteran owned critical power and cooling specialist, registered with Supply Nation and ICN Gateway. We work on both the mechanical and electrical halves of the room because in a data hall they are one system. Cooling capacity you cannot power is capacity you do not have, and protected IT load with unprotected cooling plant is a shorter outage path than no UPS at all.
Size the room by IT load, not by floor area
The single most common error in Australian server room and data hall design is sizing cooling from square metres. Floor area tells you almost nothing. The number that matters is the IT load in kilowatts, because effectively all of the electrical energy delivered to the racks leaves as heat, and the cooling plant has to reject that heat plus the plant losses and the building envelope gain. A 60 square metre room with 15 kW of IT load and a 60 square metre room with 120 kW of IT load are two entirely different engineering problems.
| IT load | Facility type | Cooling architecture | Typical UPS position |
|---|---|---|---|
| Under 20 kW | Comms room, branch office, switchroom | One or two wall-mount or upflow DX precision units. Comfort split systems are not suitable. | Single-phase rack UPS, 3 to 20 kVA, N or N+1 |
| 20 to 100 kW | Enterprise server room, regional government site | Two or three DX CRAC units in N+1, cold aisle containment, blanking panels throughout. | Single or three-phase UPS to 100 kVA, N+1 |
| 100 to 500 kW | Enterprise data hall, small colocation suite | CRAC or CRAH in N+1 with containment. Chilled water becomes economic here, condenser water where a cooling tower exists. | Three-phase modular UPS, N+1 modules, generator behind |
| 500 kW to 2 MW | Colocation hall, major enterprise or government facility | Chilled water CRAH on a concurrently maintainable ring main, dual headers, free cooling chillers. | 2N or block redundant three-phase UPS, dual A and B distribution |
| Above 2 MW | Wholesale colocation, hyperscale hall | Chilled water with economiser, dedicated plant rooms, high-density zones on in-row or liquid. | 2N throughout, compartmentalised electrical rooms |
Two adjustments almost always apply on an Australian retrofit. First, add the plant and lighting load inside the room, which on older halls can be 10 to 15 per cent on top of the IT figure. Second, design the condenser side to a realistic ambient. A condenser sized to a 35 degree design day in Brisbane will be in trouble on a 42 degree day, and the day you lose capacity is exactly the day the room is hardest to cool.
Redundancy topologies, and what they actually buy you
Redundancy language gets used loosely, particularly in tender responses. The distinction that matters commercially is not how many spare units exist. It is whether you can take any single item of plant out of service, on a Tuesday morning, with the load still running and no elevated risk. That property is called concurrent maintainability, and it is a different thing from fault tolerance, which is whether an unplanned single failure takes anything down.
| Topology | What is installed | Survives | Where it fits in Australia |
|---|---|---|---|
| N | Exactly the capacity the load needs, no spare | Nothing. Any failure or any service visit is an outage or a risk window. | Small comms rooms and test environments only |
| N+1 | One spare unit above the required capacity | One failure, or one planned service, but not both at once | The workhorse. Most enterprise halls and server rooms |
| N+2 | Two spare units above the required capacity | One failure while one unit is already out for maintenance | Sites with long parts lead times or remote locations |
| 2N | Two complete independent systems, A and B, each able to carry the whole load | Loss of an entire distribution path, including the switchgear and cabling | Colocation, banking, defence and health facilities |
| Block redundant | Several N blocks plus one shared catcher block on static transfer switches | Loss of any one block, at close to 2N availability for materially less capital | Larger multi-hall sites where full 2N capital is not justifiable |
N+1 is a capacity statement, not a design
N+1 tells you how many units there are. It says nothing about whether they are fed from separate boards, whether the chilled water reaches them through one header or two, or whether the control system will actually bring the standby unit on when the duty unit fails. We regularly survey rooms with four CRAC units described as N+1 where all four sit on a single distribution board, or where the standby unit has been in fault for months and nobody noticed because the other three were carrying the load with the alarm relay disconnected. Test the failover, do not assume it.
Redundancy levels and Uptime Institute tiers are not the same axis
A common shortcut in Australian marketing material maps Tier III directly onto 2N and Tier IV onto 2N+1. That is not how the Uptime Institute Tier Standard works, and buyers get caught out by it. Tier III is defined by concurrent maintainability, meaning multiple distribution paths with one active, and every capacity component removable for service without affecting the IT load. It is commonly delivered with N+1 capacity, not 2N. Tier IV is defined by fault tolerance and compartmentalisation, meaning two simultaneously active paths and physical separation, which does usually imply 2N or better. Redundancy count is an input to a tier, not a synonym for one. If a specification says 2N, ask separately whether it is concurrently maintainable, because the two claims are independent and only one of them protects your maintenance window.
Cooling redundancy is usually the weaker half
Electrical redundancy is well understood and well funded. Mechanical redundancy is where audits find the gaps. Common single points of failure include a shared chilled water pipework header, one condenser water pump serving units nominally in N+1, a single glycol loop, one BMS controller sequencing the whole plant, and a single power feed to a group of CRAC units. Any of these turns a stated N+1 into a real N. If cooling redundancy is a contractual commitment, it needs to be verified against the pipework and single line drawings, not against the equipment count.
Put the cooling plant on protected power. A conventional 5 kW to 8 kW rack gives you a few minutes of thermal inertia after air movement stops. At 20 kW and above that window closes quickly, and the room will hit over-temperature shutdown well before the UPS batteries are exhausted. Fans, pumps and CRAC units on generator alone are not enough, because the generator start and transfer sequence is exactly the moment airflow stops.
CRAC, CRAH and the room-level cooling decision
At room level there are three mainstream precision cooling topologies, and the right one is decided by what heat rejection the building already has, not by preference.
DX, direct expansion
A self-contained refrigeration circuit with an external condenser. The simplest to install, the easiest to make redundant unit by unit, and by far the most common choice in Australian rooms under about 300 kW. Each unit is independent, so a failure is contained. The trade-off is efficiency at scale and a condenser that has to be sized for local ambient extremes. See our DX CRAC servicing for maintenance detail.
CHW, chilled water CRAH
The room unit is a coil and a fan, with the refrigeration done centrally by chillers. Efficient at scale, far quieter in the hall, and the natural fit above roughly 500 kW. The catch is that the chilled water plant becomes a shared dependency, so the redundancy question moves upstream to the chillers, the pumps and the pipework topology. Our CHW CRAC maintenance covers the room-side scope.
CDW, condenser water
A water-cooled compressor in the room unit rejecting to a cooling tower loop. It suits buildings that already run a condenser water system, often a mixed-use tower with a data hall on one floor. Efficient, but it inherits the base building tower and its maintenance regime, which is a real risk when the data hall and the base building have different owners. See CDW CRAC servicing, or the DX, CHW and CDW comparison if you are still choosing.
Above roughly 20 kW per rack the room-level answer starts to run out and the conversation moves to in-row, rear door and liquid cooling. That is a different design problem with different economics, and we cover it in detail on our AI, GPU and high-density data centre cooling page.
PUE, NABERS and the numbers you are measured on
Power usage effectiveness is total facility power divided by IT equipment power. A PUE of 1.0 is the theoretical floor. NEXTDC's M1 facility in Melbourne became the first Australian data centre infrastructure facility certified at NABERS 5 star, running a PUE around 1.3. Older Australian enterprise halls commonly sit between 1.8 and 2.5, which means that for every kilowatt reaching a server, another 0.8 to 1.5 kilowatts is being spent on cooling, distribution losses and lighting.
NABERS for Data Centres is the Australian rating scheme, and it has three streams: IT Equipment, for tenants who control their hardware but not the building services, Infrastructure, for owners and operators of the facility services, and Whole Facility, where one party controls both. It appears in government and enterprise tenders with increasing frequency, and it is worth knowing which stream applies to you before a tender asks.
The measures that actually move PUE in an existing Australian hall, roughly in order of return on spend, are: blanking panels and floor grommets to stop bypass air, hot or cold aisle containment, raising the supply air setpoint into the upper part of the ASHRAE TC 9.9 recommended envelope of 18 to 27 degrees Celsius at the server inlet, variable speed drives on CRAC fans, correcting oversized or fighting humidity control where one unit humidifies while another dehumidifies, and finally free cooling or economiser plant. Containment and setpoint alone frequently deliver a 10 to 20 per cent reduction in cooling energy with no plant replacement at all.
Fighting humidity control deserves a specific mention because it is so common and so expensive. In a room with several DX CRAC units on independent controls, it is routine to find one unit calling for humidification while its neighbour dehumidifies, both running compressors and heaters against each other around the clock. Networking the units to a common control, or simply widening the humidity deadband, removes the conflict.
Monitoring, DCIM and BMS integration
A data hall you cannot see into is a data hall you cannot operate. Every UPS and CRAC unit we install is commissioned with monitoring integrated into whatever the site already runs, over Modbus RTU or TCP, SNMP or BACnet, into a BMS or a DCIM platform. The points that matter most in practice are: per-unit cooling status and mode, supply and return air temperature at each unit, rack inlet temperature at the top of the worst rack rather than the room average, chilled water or condenser water flow and temperature, UPS load percentage and battery health, and alarm state on every redundant unit including the ones currently on standby.
That last point is the one that gets sites into trouble. Standby plant fails silently. If the alarm relay on the standby CRAC is not wired through to the BMS, or the DCIM point exists but nobody has an escalation attached to it, then the redundancy you are paying for is theoretical until the day it is needed. Environmental monitoring should also cover leak detection under the floor and around any water path, and gas detection where flooded or lithium batteries are present. See gas detection for battery rooms and data centres.
Australian standards and compliance
Electrical installation and switchboard work is carried out to AS/NZS 3000, the Wiring Rules, with licensed electrical work and documented test results at handover. Refrigerant handling requires ARCtick licensing under the Ozone Protection and Synthetic Greenhouse Gas Management Regulations, and refrigerant logs are part of a compliant maintenance record. Mechanical ventilation and air handling hygiene falls under AS/NZS 3666, which is directly relevant to cooling coils, drain trays and any condenser water system where Legionella control applies. Building work sits under the National Construction Code, and where a data hall sits inside a leased tower, base building fire and mechanical services interfaces need to be confirmed rather than assumed.
Financial services operators carry an additional obligation. APRA CPS 230 requires regulated entities to identify critical operations, set tolerance levels for disruption, and test their ability to maintain those operations, which brings data hall power and cooling resilience directly inside the prudential scope. If that applies to you, our banking and financial services page covers the specifics.
What Indigi does for data centres and IT infrastructure
Survey and capacity position
Measured IT load, thermal survey with rack inlet readings, power quality measurement and a written statement of your real redundancy position against the drawings.
Design, supply and install
DX, chilled water and condenser water CRAC, containment, UPS and distribution. Full mechanical and electrical installation, commissioning and documented handover.
Planned maintenance
Scheduled UPS, battery and precision cooling maintenance nationally, sequenced so that redundant plant is serviced without dropping the hall below its stated topology.
Monitoring and reporting
Modbus, SNMP and BACnet integration into your BMS or DCIM, alarm escalation on standby plant, battery condition trending and refrigerant compliance records.
Indicative pricing for cooling design and installation runs from $8,000 to $15,000 for small rooms, $25,000 to $45,000 for medium installations and $45,000 to $60,000 plus for large projects, with multi-megawatt halls quoted individually because plant, pipework and switchboard work dominate the cost. UPS maintenance starts at $720 ex GST for a one-off visit or $1,940 per year ex GST on a standard annual plan. Three-phase UPS installation starts from $3,200 for small systems and $6,500 for medium systems. Single-phase UPS installation starts from $850 up to 3 kVA, $1,250 for 3 to 10 kVA and $1,800 for 10 to 20 kVA. You can estimate replacement battery spend with our UPS battery replacement cost calculator.
Frequently asked questions
Do data centres need cooling?
Yes, and continuously. Effectively all of the electrical power delivered to IT equipment is converted to heat, so a hall drawing 200 kW of IT load is producing roughly 200 kW of heat that must be rejected every second the equipment is running. Without cooling, temperatures rise fast enough that servers begin thermal throttling within minutes and shutting down shortly after. In an Australian data hall the cooling plant is also the reason the room stays inside the ASHRAE TC 9.9 recommended envelope of 18 to 27 degrees Celsius at the server inlet, which is what equipment warranties are written against.
How are data centres being cooled?
Most Australian data centres are cooled by precision air conditioning at the room perimeter, either DX CRAC units with their own refrigeration circuit or chilled water CRAH units fed from central chillers, combined with hot or cold aisle containment to stop supply and exhaust air mixing. Larger sites add economiser or free cooling plant to reduce compressor hours. High-density zones above roughly 20 kW per rack increasingly move to in-row cooling, rear door heat exchangers or direct-to-chip liquid cooling, because air alone becomes impractical at those densities.
What type of cooling is used in data centres?
Three room-level types dominate. DX, or direct expansion, is a self-contained refrigeration circuit with an external condenser, and is the most common choice under about 300 kW. CHW, chilled water, uses a room coil and fan with central chillers doing the refrigeration, and is the standard above roughly 500 kW. CDW, condenser water, uses a water-cooled compressor in the room rejecting to a cooling tower loop, which suits buildings that already run one. Beyond these, high-density facilities add in-row units, rear door heat exchangers and liquid cooling.
Can a data centre be air cooled?
Yes. The great majority of Australian data centres are entirely air cooled, and air remains the right answer for conventional enterprise racks in the 3 kW to 15 kW band. With proper containment, blanking panels and floor sealing, well-executed air cooling comfortably handles up to roughly 20 kW per rack, and in-row air cooling extends that to around 40 kW. Above that the air volumes and fan energy required stop making engineering or commercial sense, and liquid cooling becomes the practical option.
What is the biggest problem for data centres?
In Australia it is capacity constraint colliding with ageing infrastructure. A large share of the national footprint is over 20 years old, designed for 3 kW to 5 kW racks, and now carrying loads well beyond that. The result is cooling plant running at 100 per cent with no genuine redundancy left, PUE figures between 1.8 and 2.5, and electrical capacity at the meter that cannot be expanded easily in Sydney or Melbourne. Downtime cost is the consequence rather than the cause, and the Uptime Institute has reported that over 45 per cent of global outages cost more than USD 100,000 per event.
What is N+1 redundancy in a data centre?
N+1 means the facility has one more capacity unit than the load requires. If a hall needs four CRAC units to carry its heat load, N+1 installs five, so one can fail or be taken out for service with the remaining four still meeting the load. The same logic applies to UPS modules, chillers, pumps and generators. N+1 is the standard for most Australian enterprise data halls, but the count alone is not sufficient. The spare unit must be fed independently, controlled to start automatically, and tested, otherwise the redundancy exists only on the equipment schedule.
What is the difference between 2N and N+1 redundancy?
N+1 adds a single spare unit to a shared system, so one failure is covered but the distribution path, switchboard and pipework are usually still common. 2N duplicates the whole system into two independent A and B paths, each capable of carrying the full load, so an entire path can be lost including its switchgear and cabling. 2N costs substantially more and is used where downtime is not survivable, such as colocation, banking, defence and health. Block redundant sits between the two, sharing one catcher block across several N blocks on static transfer switches to approach 2N availability for less capital.
Where we work
Brisbane and QLD
Brisbane CBD, Woolloongabba, Eight Mile Plains, Port of Brisbane, Gold Coast, Sunshine Coast, Ipswich, Townsville, Cairns
National HQ at Tingalpa. High summer ambient means condenser sizing and head pressure control decide whether a hall holds capacity in February.
Sydney and NSW
Sydney CBD, Parramatta, North Ryde, Macquarie Park, Western Sydney, Newcastle, Wollongong, Canberra (ACT)
The deepest colocation market in Australia, and the one where grid capacity at the meter most often caps what a hall can host.
Melbourne and VIC
Melbourne CBD, Port Melbourne, Docklands, Dandenong, Tullamarine, Geelong, Ballarat, regional Victoria
The best free cooling hours on the mainland, and home to Australia's first NABERS 5 star rated data centre infrastructure facility.
Perth and WA
Perth CBD, Fremantle, Kalgoorlie, Pilbara, regional WA
Resources sector data halls and remote edge sites where parts lead times often justify N+2 rather than N+1.
Adelaide and SA
Adelaide CBD, Port Adelaide, Salisbury, Mount Gambier, Whyalla, regional SA
Defence, research and state government computing, served from the Melbourne hub with good economiser hours in the cooler months.
Darwin and NT
Darwin CBD, Palmerston, Katherine, Alice Springs, remote NT sites
High ambient and high humidity year round. Free cooling hours are minimal, so plant redundancy and condenser maintenance carry more weight.
Western Australia, Tasmania and Pacific Islands: we schedule data hall UPS, battery and CRAC work in planned blocks to keep mobilisation costs down, and can combine electrical and mechanical scope into a single attendance. Contact us to discuss scheduling.
Related services and equipment
- CRAC unit maintenance services across all three topologies
- CRAC installation and commissioning
- Server room cooling Australia for rooms under 100 kW
- Three phase UPS maintenance and UPS maintenance plans
- UPS battery maintenance and replacement
- AI, GPU and high-density cooling for racks above 20 kW
- Banking and financial services for APRA CPS 230 obligations
- Power integrity inspection to verify your real redundancy position
- Liebert CRAC fault codes and alarm troubleshooting
- UPS, battery and CRAC frequently asked questions
|
Need your data hall assessed or upgraded? Send us your measured IT load, rack count, existing CRAC and UPS plant, stated redundancy topology and available electrical capacity. We will come back with your real thermal and power position, an architecture recommendation and an installed price. Indigi Power and Cooling is Indigenous and Veteran owned, and registered with Supply Nation and ICN Gateway. Contact Indigi Power and Cooling CRAC Maintenance |
Partner with Us
Reliable Power & Cooling Solutions from Indigi Power & Cooling Expert supply, installation, and contracting services for critical infrastructure across Australia.