A UPS Redundancy Primer

When it comes to UPS systems, Ed Harrell has seen it all. As VP of engineering for DataSphere, which builds and upgrades data centers, Harrell has overseen the construction of no less than eight data centers in the past three years. The largest of these--a 170,000-square-foot facility for Exodus Communications--required 18 UPS modules backing up an electric current of 8.2 megawatts, or 8.2 million watts.

In other words, Harrell's in a good position to advise companies on the strengths of various UPS setups. Here's what he says about the three variations of redundant UPS systems, and what the risks or benefits of each are:

According to Harrell, many older data centers or aging corporate backup systems depend upon UPS systems that are parallel redundant. This setup features multiple UPS units, all connected and delivering power through a single switch.

The advantage of a parallel system is cost, but the risk is great: It's vulnerable because it has only a single point of failure. Harrell generally recommends that companies with parallel systems consider an upgrade. "You can take a whole system down with a single short circuit."

Isolated redundant systems are characterized by two or more systems that back each other up. If one fails, the second kicks in to cover it. As a result, isolated redundancy reduces the possibility of backup failure. This may be the most cost-effective answer for companies looking to upgrade from a parallel system without breaking the bank. Typical cost, Harrell says, is 20% to 25% higher than a parallel system.

The most luxurious of UPS setups, distributed redundant systems are the most diverse backup technology and tend to be the choice of data centers, Web-hosting companies, and co-location facilities. This technology is, essentially, a system of multiple units providing backup power to a small piece of the operation but also providing multiple secondary backups in case of any UPS component failure.

That translates to multiple points of failure, and thus reduced risk. "You reduce the impact and give more paths back to the power source," says Harrell. But he also says anyone considering a distributed redundant system should be prepared to pay as much as three times the cost of a parallel system.