Tier III vs Tier IV data centres: Uptime Tiers, TIA-942 ratings and what to specify

What do the Uptime Institute Tiers mean?

The Uptime Institute Tiers are four levels that describe how a data centre's site infrastructure, mainly power and cooling, behaves during maintenance and during failures. Each Tier builds on the one below it.

The Uptime Institute describes the Tiers as performance-based, not prescriptive. The Tier Standard states the result a design must achieve, not the equipment it must use. Any design that meets the redundancy and fault tolerance requirements is acceptable. That is why two Tier III facilities can look quite different on the drawings.

The Tiers cover site infrastructure only. The Uptime Institute says the Tier Classification System evaluates the facility, not the servers, storage, network or software inside it. A Tier IV building running a single server with a single network link still has a single point of failure.

What is the practical difference between Tier III and Tier IV?

The difference is what happens when something breaks without warning. A Tier III data centre can have any component taken out for planned work without shutting down IT. A Tier IV data centre adds fault tolerance: when equipment fails or a distribution path is interrupted, the Uptime Institute says the effects are stopped short of IT operations.

Tier IV asks for several things Tier III does not. Complementary systems and paths must be physically separated, which the Uptime Institute calls compartmentalisation, so that one fire, flood or cable fault cannot take out both sides. The Uptime Institute's technical paper on continuous cooling says Tier IV is the only Tier that requires it, so that the IT environment stays stable. The same paper recommends continuous cooling above 4 kW per rack whatever the Tier, which matters as rack densities rise. Tier IV systems must also respond to failures on their own, detecting and isolating a fault without waiting for an operator.

Both Tiers rely on dual-corded IT equipment fed from two power paths. The Uptime Institute has said that single-corded equipment in these facilities should be served by a rack-mounted transfer switch, placed as close to the equipment as possible.

• Tier III protects against planned downtime. Tier IV also protects against a single unplanned failure.
• Tier IV needs physical separation of the two sides, not only two sets of equipment.
• Tier IV needs continuous cooling. Tier III does not.
• Tier IV costs more to build, needs more space, and asks more of the operations team.

How do TIA-942 ratings differ from Uptime Tiers?

ANSI/TIA-942 is a standard, published by the Telecommunications Industry Association, for data centre telecommunications and physical infrastructure. It rates facilities from Rated-1 to Rated-4. It is not the same scheme as the Uptime Institute Tiers, and the two bodies agreed to keep them separate.

Under that agreement, TIA removed the word 'Tier' from ANSI/TIA-942 and replaced it with 'Rated'. So a 'TIA Tier III' data centre is a mixed-up description: it is either Uptime Tier III or TIA Rated-3, and the buyer should ask which. The current revision, ANSI/TIA-942-C, was released by TIA's TR-42 committee in April 2024. It covers edge data centres, higher rack power densities and liquid immersion cooling. The standard is sold by TIA, so buyers should reference it rather than quote clauses from memory.

The four TIA ratings line up in spirit with the four Tiers. TIA describes Rated-3 as having redundant capacity components and redundant distribution paths that allow concurrent maintenance, and Rated-4 as having active redundant paths and protection against single failure scenarios. The schemes still differ in scope and in who certifies.

What does it mean when a data centre is certified?

A certified data centre has been assessed by the scheme owner or its licensed assessor, not just designed to a level. A claim of 'Tier III design' or 'built to Rated-3' without a certificate is a statement of intent.

The Uptime Institute issues three certifications. Tier Certification of Design Documents (TCDD) checks the drawings. Tier Certification of Constructed Facility (TCCF) checks that the building was built as designed and performs as required. Tier Certification of Operational Sustainability (TCOS) checks how the facility is run. The Uptime Institute treats a design certification as provisional until the constructed facility is certified, because a lot can change between drawings and handover. It has also said that labels such as 'Tier III plus' have no basis in the Tiers.

TIA's certification scheme uses certification bodies licensed by TIA, which must meet ISO/IEC 17020 or 17021. TIA says a design certification is valid for one year, a constructed facility certification for three years with surveillance audits, and a ready certification for one year with yearly recertification.

Does Tier IV mean 2N, or a guaranteed uptime percentage?

No to both. The Uptime Institute says it is possible to achieve Tier IV with N+1 components, depending on how they are configured and connected to redundant distribution paths. 2N is one way to get there, not the definition.

Uptime percentages such as '99.98 per cent' are also not part of the Tiers. The Uptime Institute removed references to expected downtime per year from the Tier Standard in 2009, and says those figures were never part of the Tier definitions. A tender that asks for 'Tier III, 99.982 per cent availability' is mixing a design classification with a service level. Availability should be written as a service level in the operations contract, measured and reported, not implied by a Tier.

Which level should a public institution specify?

A public institution should specify the level each service actually needs, then pick the site design that serves the most demanding of them. Start with the list of services and how long each can be down: hospital information and imaging systems, citizen-facing portals, surveillance video, campus networks, research storage.

For many institutions the real question is whether the main data centre must be concurrently maintainable, so that UPS, generators and cooling can be serviced without an outage. That is the Tier III or Rated-3 question, and for services that run around the clock it is usually the one that matters. Tier IV or Rated-4 is justified where even a single unplanned failure in the building is unacceptable, and where the budget, space and operations team can support it.

A second site should be weighed against a higher Tier. A Tier IV building is still one building in one flood zone on one grid. A disaster recovery site at a safe distance, with tested failover, protects against events that no single-site design can absorb. The buyer should compare the cost of each option against the services it actually protects.

What should a data centre tender say about Tiers and ratings?

The tender should name the scheme, the level, the certificate expected and who pays for it. Vague wording is how buyers end up with a 'Tier III compliant' drawing and no certificate.

• The scheme and level, stated exactly: 'Uptime Institute Tier III' or 'ANSI/TIA-942 Rated-3', not 'Tier 3'
• Whether certification is required, and at which stage: design documents, constructed facility, or both
• Who engages and pays the certifying body, and what happens if certification is not achieved
• The redundancy expected for IT equipment, network and storage, since the Tiers do not cover them
• Availability as a measured service level in the operations and maintenance contract
• Operations requirements: staffing, maintenance procedures and training, which decide real uptime after handover
• Whether a disaster recovery site is in scope, and the recovery targets for each service

How Ahuva approaches data centre work

Ahuva expanded into data centres in 2024, offering end-to-end design, implementation and management. Its data centre scope covers design, server, storage, virtualisation and HCI; power, precision cooling and structured cabling; and DCIM and high-availability architectures. Its OEM authorisations include Vertiv for data-centre power and precision cooling and HPE for server, storage and compute platforms.