Redundancy considerations

Most modern pieces of IT equipment come with at least two power supply units (PSUs). Under normal operation, the PSUs are load sharing, drawing approximately equal current from the utility (electricity supply) and providing approximately equal current to the load. When two PSUs are installed, each of the individual PSUs are sized to handle the maximum load in the event of a single PSU failure or a facility electrical maintenance condition that removes input power to a PSU.

At a facility level, the different electrical dual power feeds are typically called “A-side” and “B-side” or “Left-side” and “Right-side.” These different sides are different electrical circuits that feed the IT equipment. There are several different data center redundancy architectures that provide levels of redundancy that can be categorized as N, N+1, 2N, 2(N+1), etc. It is up to the data center owner to decide on the best level of redundancy that balances availability versus cost. The most common architectures are shown in Figure 1.

There are many other possible electrical distribution architectures that are not shown in Figure 4. For example, the Uptime Institute has their Tier Classification Systems for data centers with four different electrical distribution schemes.

The most important thing to understand about dual redundancy is that measurements under normal operating conditions result in approximately one half of the capacity being used on the “A-side” and “B-side.” When determining electrical capacity for adding, rearranging, re-balancing, etc. IT equipment, carefully planning is required to avoid overloading a circuit when redundancy is unavailable.