- Sr. Building Systems Engineer »
- Senior Director of Facilities »
- Facilities Director »
- Construction engineer, U.S. Dept. of State »
- ELECTRICIAN »
Ensuring Adequate UPS Redundancy
February 12, 2014 - Contact FacilitiesNet Editorial Staff »
One way to help ensure your UPS performance meets demands is to ensure you have the right level of redundancy. While there are a variety of different UPS systems and configurations to choose from, selection ultimately boils down to the desired level of redundancy and fault tolerance of a data center. If "N" is the capacity needed by the critical load, which configuration is appropriate for the mission critical environment in question?
1N (no redundancy)
N+1 (parallel partial redundancy)
2N redundancy (distributed redundancy)
2(N+1) (parallel distributed redundancy)
Although each subsequent level of redundancy represents additional reliability and fault tolerance, it also increases cost and complexity. As a result, a UPS solution for a data center that processes millions of dollars of transactions for a bank per hour can be considerably different from one that supports generic commercial office operations during normal business hours.
In the vast majority of UPS installations, a simple 1N solution is perfectly adequate for the task at hand. That being said, failures can and do happen. In preparation for a potential failure, the question becomes: What amount of downtime in a year can be tolerated or what type of availability is required? Availability, or the percentage of time that the UPS system is functional, introduces the variable of mean time to repair (MTTR), as elements like unscheduled maintenance can also affect uptime. Therefore, the value of business continuity also needs to be taken into consideration when determining the required level of availability. The following equation can be helpful:
AVAILABITY = MTBF/(MTBF + MTTR)
So, with an MTBF (mean time between failure) of 100,000 hours and an MTTR of 4 hours, we can get 99.996 percent availability out of a UPS. This equates to 21 minutes of downtime a year. But 21 minutes is a purely statistical calculation. If there is a failure, the MTTR still means 4 hours of downtime before repair. And that MTTR doesn't account for the availability of parts or the response time of the service technician.