masterzphotofo - Fotolia
Data center power infrastructure essentials prevent downtime
Power infrastructure upkeep is essential to reduce damage in an unplanned outage. UPS and generator maintenance, plus disaster planning, should be a part of regular operations.
A sudden loss of power can bring an entire data center operation to its knees. Power outages are harmful to IT systems because an outage and can result in lost data, corrupt files and damaged equipment. An on-premises data center requires a backup system that is integrated directly into the power infrastructure to ensure that critical systems stay online even if the lights go out.
Data center power infrastructure should support two modes of operation: regular and emergency. Under regular operations, a data center taps into the main power grid and consumes electricity from one or more utility companies. Some organizations might use generators to supplement their power needs.
Generators, however, often serve another vital purpose in the power infrastructure: They provide backup electricity during emergency operations. In this scenario, one or more generators supply the data center with its power, delivering energy as alternating current (AC), just like the main power grid.
The importance of UPS hardware
A data center typically deploys one or more automated transfer switches to switch between primary and secondary power sources. If the primary power source becomes unavailable, the transfer switch transmits a startup command to the connected generator and then switches over to secondary power when the generator starts producing electricity. Once primary power is restored, the transfer switch changes back to primary power and transmits a shutdown command to the generator.
A data center's power infrastructure includes one or more uninterruptable power supply (UPS) appliances. The UPS serves two crucial purposes: to protect equipment from voltage spikes and provide short-term power in an outage.
Another common power infrastructure component is the power distribution unit (PDU), which is a high-grade power outlet that receives power from a UPS and then distributes it to IT systems. A PDU does not generate power or protect against power spikes, which is why it's normally used in conjunction with a UPS.
Power regulation and routing is done through switchgear or main distribution boards, as well as transformers. This ensures that the power supply is the right voltage and current. The exact equipment configuration depends on the data center's size, number and type of IT systems.
Ensure continuous operations
Data center power infrastructure must support facility systems such as lighting, alarms, sensors, fire monitors, cooling units and dehumidifiers. The infrastructure must also provide electricity for IT systems such as servers, storage devices and network components so they can operate without interruption, even if the power goes out.
Power that targets IT systems runs through the UPS continuously during regular operations. The UPS ensures that connected batteries remain fully charged and ready to support emergency operations. The batteries provide enough power to keep IT systems -- and perhaps some facility systems -- running for a short time, as little as five minutes, depending on the number and types of batteries.
Some UPSes use flywheels or supercapacitors rather than batteries. Flywheels employ accelerating rotors to store kinetic energy that is convertible to electricity; supercapacitors utilize electrostatics to provide high-density energy storage as needed.
A UPS falls into one of three categories: offline, line-interactive or online. If the primary power source goes out, the UPS switches to battery power, converts it to AC and sends it on to the PDUs. A line-interactive UPS works much the same way, except that it conditions the primary power as it passes through the appliance to prevent voltage spikes.
An online UPS converts electricity to DC power and charges the batteries. Any remaining power is conditioned and converted back to AC for PDU output. Large data centers that run critical workloads often implement online UPS appliances because they provide the highest level of protection, though they can increase operation costs.
Regardless of the UPS type or how an organization stores and designs backup power, the goal is the same: to provide enough time to shut down IT systems or to get the backup generators running and producing electricity. Ideally, the generators will be online after less than a minute of a detected failure, allowing the UPS systems to return to normal operations.
Backup generators generally run on diesel or gasoline. Most data centers have enough fuel on hand to run systems for 24 to 48 hours. The number of generators and total required voltage depend on the data center's specific power requirements.
Organizations must ensure that their generators operate safely, meet applicable environmental regulations and have admins monitor for exhaust fumes such as carbon monoxide or nitric oxide.
Proactively plan for power outages
Every data center should have a disaster recovery plan in place that outlines what steps to take and what role each person plays in the event of a power outage; following this plan is essential during an outage.
This means shutting down systems in a prescribed order, ensuring that generators are running and properly ventilated, monitoring temperatures to prevent overheating and verifying that emergency systems such as pumping equipment is operational.
There must also be a plan once power is restored. Systems should be recalibrated in their prescribed order and tested to confirm everything is running as expected. Emergency equipment should be shut down and prepared for the next crisis. Above all, participants should maintain proper communications both during and after a power failure, keeping anyone affected by the outage updated on infrastructure status.
Regular equipment maintenance and testing is also necessary for data center power infrastructure. A UPS with bad batteries or a generator with contaminated fuel is little help to the data center when the lights go out and it can affect troubleshooting time.
