The price of datacentre reliability

Building a datacentre is one of the most expensive tasks an IT director can face. Understanding where the costs and benefits lie is a key first step to success


Building a datacentre is one of the most expensive tasks an IT director can face. Understanding where the costs and benefits lie is a key first step to success

Building a datacentre is one of the largest capital investments for enterprise IT organisations. Costs can be tens or hundreds of millions of pounds, depending on size, location, and most importantly, levels of reliability and initial power rating.

Gartner surveyed four engineering/consulting firms that submitted project data for an more than 30 data¬centre projects. The projects were completed in the past two years and classified into four tiers, according to definitions from the Uptime Institute.

Tier one is the lowest level and tier four is the highest, with complete active dual-path electrical distribution, power generation and uninterruptible power supply (UPS) systems. Tier one specifies annual outage up to 28.8 hours; tier two specifies 22 hours; tier three specifies 1.6 hours; and tier four specifies 0.4 hours of annual outage, or 99.995% availability.

Not surprisingly, the survey revealed significant increases in cost levels since 2002. In most cases, project costs increased by 60% to 80% for tier two and tier three projects.

Increases can be attributed to three basic cost drivers. First, there has been an increased emphasis on reliability. This relates to the need for redundant electrical and mechanical systems to eliminate single points of failure.

The higher the tier, the greater the level of redundancy and, consequently, the more extensive the investment costs for duplicate electrical and mechanical systems, including UPS systems, generators, multiple air-conditioning units, and the attendant duplication of electrical cabling and piping for dual water distribution.

Second, the survey found an increased initial power rating associated with higher-density computer equipment, such as blade servers, storage area network equipment and disc arrays. Not only is increased power required for the more dense IT equipment, but also higher power is required for additional air-conditioning systems to cool the incremental heat load.

Finally, there was increased investment to harden the datacentre facility’s structure in response to geographic or terrorist risks.

Tier levels will drive design specifications for new datacentres. The higher the tier, the higher the investment for building construction and mechanical/electrical equipment. Tier level or degree of reliability will be determined by the criticality of datacentre operations.

In the financial industry, high availability and reliability (typically tier three and tier four) are required to support 24x7 financial transaction activities and funds exchanges. For other enterprises a lower level of reliability may be acceptable.

The Uptime Institute uses two key definitions in establishing level of reliability. The first is the concept of “concurrent maintainability”. The institute defines the acid test for concurrent maintenance as “the ability to accommodate any work activity without disruption to computer room processing”. Tier one and tier two do not meet this test, but tier three does.

The second concept is “fault tolerance”. The Uptime Institute defines the acid test for fault tolerance as “the ability to sustain an unplanned failure or operations error without disrupting computer room processing”.

Tier three does not meet these criteria, but tier four does. Therefore, the incremental investment for dual electrical and mechanical systems to meet concurrently maintainable and fault-tolerant criteria underlies the significant increase in capital investment from tier three to tier four. One cautionary note: although a higher tier level can address infrastructure reliability, human error and operational issues can also affect availability.

Electrical costs are the single largest expense in the datacentre (42.4%). These include the cost of UPS systems, mechanical generators, power distribution units, transfer switches, electrical cabling and conduits, as well as installation

The second highest cost (21.2%) is for mechanical systems such as air conditioning. This includes cooling towers, computer room air-conditioning units, plumbing and associated installation labour. 

Gartner’s recommendation is to focus on the electrical and mechanical components of the project budget; these costs should range between 50% and 70% of the budget, depending on the tier level and power rating. It is useful to develop a detailed computer equipment plan that will inform total power requirements.

Gartner also advises the use of a design of 50 watts to 100 watts per square foot. Aim to come in at the lower end of this range initially, and then plan to add capacity on a modular basis during the life of the facility: usually 10 years. 

Finally, planning should anticipate the possibility of water cooling in the future, so provisions should be considered for a chilled water loop throughout the raised floor area.


Tier one
● Single path for power and cooling distribution
● Less than 28.8 hours of downtime per year

Tier two
● Single path for power and cooling distribution
● Redundant components
● Less than 22 hours of downtime per year

Tier three
● Multiple power and cooling distribution paths, but only one path active
● Redundant components
● Concurrently maintainable
● Less than 1.6 hours of downtime per year

Tier four
● Multiple active power and cooling distribution paths
● Redundant components
● Fault-tolerant
● Less than 0.4 hour of downtime per year

Michael Bell is research vice-president and project co-director at Gartner

Read more on Data centre hardware