Update 4/10/2014 - The TIA will remove "tiers" from its data center benchmarking.
Many data centers advertise themselves as a specific Tier, based on a scale from I – IV. But these classes are generally poorly defined and in many cases misused. Case in point: Green House Data recently exhibited at an industry event in Denver. A man walked up to the table and started asking about the company’s data centers.
“You guys are up in Cheyenne, right?” he said, “What kind of facility? Tier II? Tier III?”
“We meet the standards for Tier III, yes,” I replied. “99.9% availability.”
“Raised floors?” he continued. I shook my head.
“It has to have raised floors to be Tier III,” the man said with certainty.
In actuality, raised floors are only required in some circles. Different accreditations use extremely similar tiers, with several key differences. This blog aims to set the record straight, spelling out what is included in each tier from two major accreditation organizations.
The two main data center tier classifications were developed by the Telecommunications Industry Association (TIA) and the Uptime Institute (UI). Each uses Tier I, II, III and IV to describe various levels of reliability and infrastructure design in the data center, but they diverge in several key ways. The basic idea, however, remains the same: an enterprise-class data center must be designed with redundant components and backup systems in order to avoid downtime, even during maintenance periods.
The Uptime Institute consists of a group of companies who perform IT services and consulting for the data center industry. They first created their Tier system in 1995. TIA was formed in 1988, born from other telecommunications industry organizations stretching back to 1923. They released their Telecommunications Infrastructure Standards for Data Centers in 2005. Both standards stipulate some other requirements for cabling, floor layout and more, but the main points align nicely:
ClassTIA-942 StandardUptime Institute Standard Tier I
- 99.671% availability
- susceptible to disruptions from planned or unplanned activity
- single path for power and cooling distribution
- N+0 – no redundancy
- May or may not have a raised floor, UPS or generator
- Annual downtime of 28.8 hours
- Must be completely shut down for maintenance
- Susceptible to disruptions from planned or unplanned activity
- Single path for power and cooling distribution
- N+0 – no redundancy
- Includes a generator and UPS for outages and power spikes, with 12 hours of generator fuel minimum
- Must be completely shut down for maintenance
Tier II
- 99.741% availability
- Less susceptible to disruption from planned/unplanned activity
- Single path for power and cooling
- N+1 – includes redundant components
- Includes raised floor, UPS and generator
- Annual downtime of 22 hours
- Maintenance of power path and backbone may require shutdown
- Less susceptible to disruption from planned/unplanned activity
- Single path for power and cooling distribution
- N+1 components including generators, UPS, energy storage, chillers, heat rejection, pumps, cooling and fuel tanks
- Includes UPS and generator with 12 hours of fuel
- Redundant components can be removed for maintenance without disruption, but distribution path maintenance may require shutdown
Tier III -
Concurrently
Maintainable
- 99.982% availability
- Normal activity will not disrupt critical operations, but unplanned events could still cause disruption
- Multiple power and cooling distribution paths with only one active at one time
- N+1 redundancy
- Annual downtime of 1.6 hours
- Includes raised floor and ability to maintain full operation while performing maintenance on power path or backbone
- Normal activity will not disrupt critical operations, but unplanned activity/human error may
- Multiple distribution paths for power and cooling with only one active at any one time
- N+1 redundancy
- All IT equipment is dual-powered or features transfer devices
- Includes UPS and generator with 12 hours of fuel for every “N” capacity
- Can maintain full operation with any component of distribution path removed for maintenance
Tier IV -Fault Tolerant
- 99.995% availability
- Normal activity does not disrupt critical operations, can experience at least one unplanned event with no impact
- Multiple power and cooling distribution paths
- 2(N+1) redundancy – 2 UPS each with N+1
- Annual downtime of 0.4 hours
- Includes raised floor and ability to maintain full operation during maintenance
- Normal activity does not disrupt critical operations, can experience failure of any component with no impact
- Multiple power and cooling distribution paths that are independent, diverse and simultaneously active
- N+1 redundancy with physical separation
- Continuous cooling required
- UPS and generators required with 12 hours of fuel for “N” capacity
- Each and every component can be removed from service for maintenance without affecting critical systems
The two class systems look quite similar, with a main difference being those tricky raised floors, which can be a source of debate among data center designers. In the past, raised floors were the preferred method. They solved a number of issues including:
- cooling and air distribution
- management of bulky cables
- copper grounding grid
- location to run utility piping
.png){kind=logo}
