In this video find out how different types of equipment are configured to deliver reliable, redundant power for rack servers. We take a deep look into how power from a utility company gets delivered to a data center and its power distribution chain in the North American.
Welcome to this 3rd video of the Rack Power Course. In this video, we're going to talk about how power from a utility company comes into a data center and the power distribution chain within the data center. To keep the discussion simple, this information will focus on North American data centers. View the international power voltages video to learn about the different voltages used in data centers around the world.
The best goal for a class 4 data center would be to have 2 feeds from separate utilities or from at least 2 main substations. This would provide 2 completely redundant power chains, including redundant access to the utility power grid.
Since this is very expensive and usually not feasible, the most popular option is to back up the utility power feed with diesel generators. A series of switches on both the utility feed and the generator feed are used to provide a level of redundancy for all the equipment in the data center.
This seemingly complicated chart is actually a simplified view of what many data center schematics look like. Let's zoom in on this and take it piece by piece. We'll start with the utility feed.
Power coming from the utility distribution substation can range from 2 to 30 plus kilovolts (kV). Since the power entering the building is generally 480 Volts, there has to be a transformer that can reduce the voltage down to 480 volts. The power then passes from the transformer through an automatic transfer switch or switches. The purpose of this switch is to sense when sufficient power is no longer coming from the utility due to brown outs, blackouts, or short term power fluctuations. If the utility power is interrupted, the automatic transfer switch will signal the backup generators to turn on. The generators will start powering up and deliver electricity to the data center equipment.
Many power disruptions can be under just 3 seconds so a data center operator will not want the generators turning on needlessly. Also it takes a little bit of time for the generators to completely power up and reach a stable level of power ouput. To bridge the gap between when the utility power stops, to when the generators are delivering full power, We have a standby power source called an uninterruptible power supply, which is known as UPS.
The standby power provided by UPS is known as ride through power and can be either a battery or a flywheel system. Both systems store specified amount of energy to span the time between utility power outage and generator power startup. The flywheel UPSes, when used in series, can generally generate sufficient power for at least a few minutes; Battery systems can provide power for longer durations of 15 minutes or more. There are benefits and drawbacks with both types of systems. After utility power is restored and stable, the automatic transfer switch switches back to the utility feed and shuts off the back-up generators. The UPS is then recharged through energy storage capabilities.
The power from the UPS is usually wired to the power distribution units, frequently referred to as PDUs. Note that these very large PDUs are not the same as the rack PDUs we'll reference later. In these videos, I'll refer to these large PDUs as floor PDUs to differentiate them from the rack PDUs.
The floor PDUs transform the power from 480 volts down to either 400 or 208 volts. Older data centers were built with 208 volt power whereas most new data centers are using 400 volt power. The reason for this development will be covered later in another video.
Floor PDUs are connected to a series of remote power panels referred to as RPPs. These are larger versions of the breaker panel in your home. If you look at this panel, it has multiple points where you can connect wires to run to your equipment. These points are called poles. There are panel boards that can have 168 poles.
If you're delivering 3 phase power to the rack, then you would connect the electrical wire to 3 separate poles. These 3 wires are within a conduit that's called a whip and these whips run under the raised floor of the data center. At the end of the whip is the outlet box.
Each rack should have two outlets from 2 different RPPs attached to 2 separate UPSes as part of the redundant power chain. The rack PDU in each rack plug into the outlets. Most servers now have 2 power supplies and it's important to plug each power supply into a different rack PDU. This way if a UPS should fail, then the connected floor PDU and RPP will fail and one rack PDU will fail. If properly configured, the other rack PDU along with the other power chain will remain operational and your servers won't go down.
An important exception to the just mentioned power chain is the increasing use of overhead bus systems. These systems connect to the UPS and eliminate the need for Floor PDUs, RPPs, and whips running underneath the raised floor. Instead the overhead bus systems have sophisticated channels running over the racks and other equipment. Outlet boxes can then be placed wherever needed along these channels and are simply connected into the overhead busway. The rack PDUs are then plugged into these outlet boxes above the racks.
This has been a very simplified review of the power chain within a data center. Many of the topics that were briefly mentioned here can be the basis for extensive discussion by themselves.
The most important take-away from this video is that to have reliable redundant power for your servers in the racks, there are many sophisticated pieces of equipment that need to be properly configured.
The next video in this series will discuss the range of voltages used in data centers around the world.