Data Centers classification

Any data center is a professional version of a server room. If you dig deeper, there are many differences: scale, cost, capacity (power), cooling, security and stability.

Building the right data center is impossible without understanding and applying sophisticated engineering and information technology. Businesses are guided by industry standards: TIA-942, TIA/EIA-568-B in the US or EN 50173-5 in Europe. If the goal is to get a predictable and reliable IT infrastructure, standards are essential.

Data Centers Classification

Since there are a huge number of tasks that data centers solve, their classification is tiered as:

Data centers by ownership and management

• Owned (corporate): Owned and operated by the organizations that use them.
• Colocation (mixed): Lease space, power, networking, and cooling to customers who want to host their equipment.
• Cloud-based (hosted): Owned and operated by cloud service providers.

Data centers by size

• Micro: Up to 10 servers. Often it is a regular server room, but with some data center technology (access control, cooling). Micro data centers are deployed on company premises and used for backup, website hosting, network management, file storage or other office operations.
• Small: 10 to 100 servers. These data centers can support medium-sized businesses or branches of large companies. Internet service providers use small data centers to sell hosting services.
• Medium: 100 to 1,000 servers. They can handle the needs of large businesses and be used by ISPs to sell cloud services. Medium data centers have quite complex infrastructure and architecture as they are required to have high availability and security.
• Large: 1,000 to 5,000 servers. Large data centers are often used by corporations or cloud providers that support a large number of users or process large amounts of data. For example, social networks, federal banks, and large E-commerce businesses.
• Hyperscale: more than 5,000 servers. These data centers are usually owned by the largest technology companies. They can serve millions of users and process exabytes of data.

Data centers by purpose

• Core: These are the main operational data centers that perform the day-to-day tasks of a company: data processing and storage, website and application hosting, backup, recovery, and more. To keep primary data centers running continuously (with high availability), equipment redundancy techniques are often used: redundant power and cooling, storage, etc.
• Backup (Backup): This is a backup plan for the primary data center. If the primary data center fails, the backup one will take over the entire load so that business processes don't stop. Backup data centers can be "hot", "warm" or "cold", depends on the budget and implementation. "Hot" backup data center is always running and ready to take on the load at any time. "Warm" is preconfigured and ready for quick startup when needed. "Cold" backup data centers need to be configured before startup - it takes longer. Backup data centers are often in a different geographic location to minimize the risk associated with local disasters such as earthquakes, floods or fires.

Data centers by mobility

• Mobile: Data centers that can be moved to a new location and quickly deployed when needed. They are usually transported by trucks, trains, ships, and even airplanes.
• Containerized (modular): Data centers that are housed in adapted shipping containers (modules). They can be moved and quickly deployed when needed, which is great for temporary tasks in remote locations.
• Fixed: These are data centers that are housed in a building or some part of it. There are 3 options here:
  • Purpose-built buildings: These are originally created to house a data center. They usually have the infrastructure in place: cooling, power, security systems, and more.
  • Retrofitted and modernized buildings: Buildings that were not built for a data center, but have been retrofitted for this purpose. It could be an old office building, industrial space, warehouse or even a bunker.
  • "Trash" Data Centers: This is the cheapest type of data center and it is placed in unsuitable buildings or spaces such as dusty warehouses or hangars. The right infrastructure and conditions for a data center are often not in place there. Risks are increased and functionality is reduced.

Data centers by type of location on the ground

• Aboveground: It is the most common type of data center, that is deployed in structures above ground.
• Underground: Underground data centers offer greater security because they are more difficult to physically penetrate or damage. For example, they are more resistant to extreme weather conditions. Usually they are mine conversions, former military bunkers, tunnels or underground storage facilities.
• Floating: They are special vessels or platforms on which the data center is deployed. These vessels may be moored to a dock, be in the open ocean, or sometimes underwater.

Data centers by fault tolerance

The Uptime Institute (UI) offers a classification system that defines four successive stages of data center reliability, where each successive level includes all aspects of the previous ones.

Large companies, such as Huawei and LinkedIn, are audited by UI to obtain Tier Certification, which is an international industry standard for designing, building and operating data centers.

Table comparing TIER levels of data processing centers

Below is a table summarizing the four data center tiers where you can see the difference.

Tier I data center

99.671% uptime, up to 1,729 minutes of downtime per year.

Tier I is the initial level of infrastructure without redundancy (according to the Uptime Institute). This is where the server room ends and the data center begins.

Tier I infrastructure provides the power and cooling capacity to support a full IT load. Power and cooling are supplied through a single channel, and there is no redundancy for critical systems. For regular maintenance or emergency repairs, staff must shut down the data center completely.

The requirements for Tier 1 data centers are as follows:

1. Uninterruptible power supply (UPS) for protection against power surges and outages.
2. A fuel-fired power generator against prolonged power outages.
3. Separate room for IT systems.
4. Active cooling of the data center, working even after hours.

Purpose: Non-critical applications.

Application: Suitable for non-critical applications such as data archiving, backup, e-mail, small websites, etc. These can also be systems that can survive a short interruption without serious consequences.

Tier II data center

99.741% uptime, up to 1361 minutes of downtime per year.

Tier II is all the features of Tier I + partial redundancy of critical components (power and cooling) which gives a higher quality of service and security in case of failures.

But Tier II data centers still have a single backbone power source — sometimes you'll have to shut down operations for maintenance and repair. So Tier II is not suitable for critical applications and services that need 24x7 availability and reliability for many years.

The requirements for Tier 2 data centers are as follows:

• Redundant fuel-fired electrical generators.
• Fuel tanks and cells.
• Energy storage devices (including batteries).
• UPS modules.
• Chillers (refrigeration units).
• Surplus refrigeration units.
• Heat removal equipment.
• False floors (raised floors).
• Pumps.

Purpose: Supports basic business processes.

Use: Suitable for basic web services and sites, back-office business systems and databases, non-critical business applications, hosting (including entry-level visualization and cloud solutions), and small game server hosting. Systems that can tolerate minor interruptions, but for which a prolonged outage is a problem.

Tier III data center

99.982% uptime, up to 95 minutes of downtime per year.

A Tier III data center has redundant components and independent power feeds, networks, etc. to support the IT load, allowing maintenance without interruption. This differentiates it from Tier I and Tier II, where maintenance may require a shutdown. Tier III additions to Tier II components ensure that any piece of equipment can be shut down without impacting business processes.

The requirements for Tier 3 data centers are as follows:

• All Tier 1 and Tier 2 components + N+1 redundancy strategy. For example, if you have three servers working together to handle the load (N=3) and you add one more server as a backup (+1), you get a "3+1" configuration.
• Tier 3 requires a solution that can maintain operation for a minimum of 72 hours in the event of a local or regional power outage.

Keep in mind that the N+1 strategy will protect against single hardware failures, but if two or more critical components fail at the same time, the entire IT infrastructure will fail.

Purpose: Critical business applications and systems.

Use: Suitable for more critical business applications that require high availability and reliability, such as CRM systems, e-commerce, cloud-based middleware solutions, and large game server hosting. This tier is also suitable for companies for which an outage could result in significant financial losses.

Tier IV data center

99.995% uptime, up to 26 minutes of downtime per year.

Like Tier III, Tier IV has redundant components and independent power feeds, networks, etc. to support IT load, but unlike previous tiers, Tier IV provides full redundancy and standalone capability, allowing maintenance and repair without interruption.

The requirements for Tier 4 data centers are as follows:

• All Tier I, Tier II and Tier III components + double redundancy (2N+1). This means that every critical element of the system is backed up, as well as additional equipment in case of failure. For example, if you have four servers working together to handle the load (N=4), and you add four more servers as backup (+N), and one more as additional backup (+1), you get a "4+4+1" configuration.
• Each power channel is independent, so the failure of one does not cause a domino effect for the other components.
• After a local or regional power outage, operation must continue for a minimum of 96 hours.

Purpose: Vital to an organization's mission or goals.

Use: Designed for mission-critical applications and services that require maximum availability and reliability. These may include large cloud providers, financial institutions, major online gaming, social media, and other services for which an outage is unacceptable.