Power is one of the most important elements for a data center, and if it goes out, it can cause disasters in many companies, for example, the 1998 Eastern Canada and Northeastern United States ice storm, and the 2003 Eastern United States. Given the tremendous load and increasing demand imposed on our electrical power infrastructure, more serious outages can be expected in the future. Because of this potential threat, businesses that are heavily dependent on their IT infrastructure may want to seek ways to increase the availability of their server room or
Uninterruptible power supply
The uninterruptible power supply (UPS) is the most basic and most common form of backup power. It relies on a string of batteries to provide instantaneous but temporary power to systems connected to it in the event of a power outage or even sudden power drop or sag (also called brownout). The number or charge capacity of the batteries (battery runtime) will dictate how long a UPS can provide power. The battery runtime is usually calculated in minutes, and is meant to give enough time for another source of power to come online (e.g., emergency generator) or allow for the graceful shutdown of IT systems to prevent data corruption.
Many IT managers are tempted to increase the UPS battery runtime by adding enough batteries to support the IT power load for over an hour in an effort to ride out a power failure, which is typically a short duration depending on the cause. While increasing the UPS battery runtime may seem like a good idea, it is a waste of money in many cases. Cooling systems are not connected to a UPS, and unless they benefit from an alternate source of power like a generator, they will cease to function during an outage. In most data storage environments, the computer room will quickly overheat and cause systems to fail or shut down long before the extended UPS battery capacity is exhausted.
Emergency power generators
Emergency power generators are usually diesel- or natural gas-powered engines that spin a generator to produce electricity. Emergency power generators are usually combined with an Automatic Transfer Switch (ATS) that detects when utility power is out and triggers the emergency power generator startup. These installations are complemented by UPS to provide temporary power to the IT systems while until the generator starts up (normally less than a minute).
But many people may ask, "What about power supply to the rest of the office?" Data centers usually have a generator with only enough capacity to support the IT equipment and cooling system in the event of a power outage disaster. This is not a problem if the data center is at a different location from the main company building. That way, IT users and customers can access the systems from remote locations. But in cases where users are in the same building, having a fully powered data centers is not much help if the users are in the dark or without power for their workstations. In such cases, that generator was a waste of money.
Redundant power distribution
The Uptime Institute developed a tier system to rate the availability of data center infrastructures. The availability rating ranges from Tier I for non redundant power distribution paths to Tier IV for fully redundant, concurrently active, dual power distribution paths with an expected availability of 99.99%. This means that redundant power and active dual power paths from the utility power feed power to the building all the way to dual power cords and power supplies in IT systems.
In fact, the informative annexes included with the Telecommunications Industry Association's standard for data centers (TIA-942) take power redundancy to the next level by recommending that the data center "should have at least two utility feeders from different utility substations for redundancy."
Knowing this, in the event of a major fire in the building's utility room, the fire department would likely shut off power for the entire building and require that the emergency power generators not be started until they have put out the fire. In this situation, a costly power infrastructure could be useless.
The bottom line in data center disaster recovery planning
Overall, there are several instances where highly available and redundant power is required; hospitals and telecommunications facilities than cannot be relocated during a power failure are prime examples. That said, most IT managers must review the criticality of their systems before making that decision. If all the systems in their data storage environment are critical to a point that they cannot sustain any kind of outage due to a power failure, then they are also likely already all configured as high-availability (HA) clusters with data mirroring to avoid other types of outage. In that case, power redundancy is the next logical step. On the other hand, if only a limited number of systems are highly critical, failover to a remote site may be a more cost-effective solution than fully redundant power.
We must always keep in mind that regardless how redundant and highly available a single data center facility was designed, the location itself can still be a single point of failure.
About this author: Pierre Dorion is the data center practice director and a senior consultant with Long View Systems Inc. in Phoenix, Ariz., specializing in the areas of business continuity and DR planning services and corporate data protection.
This was first published in April 2010