Sunday, September 2, 2007


General remarks about Tier classifications

This photo was taken at one of Google’s data centers.

This is Part-3 about the subject of Tier classification. For Part-4, click here. A new tab or window will open.

To rewind to Part-1, click here. A new tab or window will open.

Would you like to learn the business factors that should be considered in selecting a Tier? Click here to read it. A new tab or window will open.

The tier of a Data Center is determined by the rating of its weakest system. For example, a center with a Tier-4 power configuration that has a Tier-3 cooling subsystem will yield a Tier-3 classification.

The center’s rank is always the lowest of its individual subsystems. It is not the average of the rank of its individual subsystems.

The MTBF (Mean Time Between Failure) of a center’s subsystems is irrelevant in determining the center’s tier. So is the number of components or systems. Here are two examples. The first example compares how a backup set of duplicate chillers can be added to a site. The backup set can be:
  • arranged along the same cooling distribution paths, or
  • arranged along a second, independent path.
The first arrangement fulfills a Tier-2 center’s criteria. The second arrangement fulfills a Tier-3 center’s criteria.

The second example compares how two separate in-line UPS batteries are controlled. The batteries can be controlled by:
  • a common input or output switch, or
  • separate independent switches.
Once again, the first arrangement fulfills a Tier-2 center’s criteria and the second arrangement fulfills a Tier-3 center’s criteria. UPS configurations that share the same input or output switch gear almost always require the server room to shut down for routine maintenance. In addition, a common switch creates another single point-of-failure. A Tier-2 center has a duplicate set of critical electrical and cooling equipment. A Tier-3 center has a duplicate set of all components and equipment that supports IT operations. The connectivity, electrical, and mechanical systems, for example, must be in duplicate.

Sites that will be built from the ground up should be designed for a future higher tier level, or at the very least, designed to anticipate future power requirements. The owner should take advantage of the relatively small cost difference between a Tier-3 and -4 infrastructure before the facility is built.

A Tier-4 Data Center can be summarily described as being fault-tolerant and concurrently maintainable. A Tier-4 Data Center will theoretically never go down regardless of the failure of any of its subsystems.

Personnel operations—the main factor in determining sustainability—play the biggest role in the uptime of a Data Center. For that reason tier ranking can only be performed objectively through the center’s topology, architecture, and components. Sustainability factors directly or indirectly account for 70% of all downtime. The performance of individual data centers within the same tier largely depend upon sustainability. The correct implementation of sustainability factors decrease the cost and risk of completing maintenance or hasten the recovery of the center from disruptions.


Sustainability largely determines the uptime performance of individual data centers within the same tier. Sustainability factors make the difference between an easy maintenance procedure or a difficult one; between an inexpensive or costly one; and a convenient or awkward one. A difficult, costly, or awkward maintenance procedure increases the chance that it will be delayed or skipped. And missed maintenance increases the chance of equipment or component failure. This section provides examples of infrastructure characteristics that impact sustainability. These characteristics are details of design, IT architecture, or implementation.
  1. The ability to switch the power source of all mechanical components so they continue running before starting maintenance work on an electrical panel.
  2. The placement of a critical component in a cramped area when it could have been placed elsewhere.
  3. The placement of engine power generators and switching gear inside the facility instead of outside will eliminate the effects of weather.
  4. The decision to limit the aggregate load on any subsystem to 90% of rated capacity instead of 100% will improve stability and prolong equipment life.
  5. Compartmentalization refers to the physical separation of the primary and secondary paths. Tier-4 sites have compartmentalized subsystems. Personnel can attack a fire in the primary path’s area if it's physically separated from the secondary path.

Sphere: Related Content

No comments: