But the answer is not always so simple: different users have different requirements for reliability and uptime; there are a variety of ways to achieve different levels of power redundancy — and redundancy costs money.
AIS can help you determine exactly what type of power redundancy will best mitigate your particular risks at the lowest overall cost. But while each client is different, there are a number of common questions and considerations about power redundancy that apply to all. This article will help you explore your requirements and options.
First, there are a number of considerations that you should keep in mind when thinking about your power redundancy options:
- What is my business model? Do I need to remain up 24/7? If not, how much downtime can I accept, both average and in any one instance? (An average of 1% downtime might seem acceptable, but that’s some 87 hours across the course of a year. If that downtime occurs all at once, it can devastate your business.)
- How soon do I need to be back up when the inevitable downtime happens? Immediately, with no loss of service? A few seconds? Minutes? Hours or days?
- What are my resource constraints? Consider not just how much redundancy will cost, but also your system architecture. Do you outsource processing and storage to a “cloud”? Do you manage your own system?
And remember that power can fail at any point in the chain:
- Failure in the feed from your electrical utility company. This can be mitigated through the use of dual utility feeds or emergency generators, and is the subject of another paper.
- Failure at a point within your own power distribution network.
- Failure at the individual device’s power supply.
- Power failures occur not only because something breaks, but can also happen as the result of routine maintenance. Redundancy allows you to take elements of the power chain offline so they can be maintained or replaced without interrupting operations.
Some Power Distribution Fundamentals
Almost all IT devices (servers, storage devices, switches, etc.) are powered by direct current (DC), while the power outlet that supplies the power is alternating current (AC). The devices rely on a power supply within the device that converts AC power to DC, and virtually all IT devices are configured to receive their power from dual power supplies.
Under normal conditions, each of the two power supplies provides half the power the device needs, but when one fails, the second can take up the entire load without interruption. Redundancy at the device level can also be achieved through a static transfer switch instead of two power supplies in each device, but the switch does not protect against failure of the power supply itself.
Static transfer switches are a good option if downtime in individual devices is not a critical problem. All of the above means that at the device level it is a relatively simple task to configure the system for power redundancy. The complications begin further up the chain.
The power supplies in the individual devices are fed by power distribution units (PDUs). The PDUs distribute power from the source, typically an uninterruptable power system (UPS) to electrical panels that supply the individual devices. The majority of power failures occur at the device’s power supply, at the PDU, or at the panel. By making these redundant, you can avoid downtime for most power failures.
The A + B Power Option
The A + B power option can provide 99.9% (“three nines”) reliability, or about nine hours of downtime over the course of a year. This option provides redundancy from the utility feed down to the individual devices.
With the addition of emergency generators or a dual power feed from the utility, the A + B option can achieve close to 99.99% reliability, or about two to three hours of downtime per year, averaged over five years. (With the A + B option, most downtime will be from power outages at the utility company, which may only happen once every few years, but can take many hours for the power company to resolve. So remember that the two to three hours per year is an average; individual outages may be longer.)
For many businesses, the A + B option is the best choice.
Some Common Misconceptions
IT managers often share a number of misconceptions about power redundancy:
- Redundant power is too expensive. Yes, redundancy does cost money to implement. But it is usually less expensive than suffering extended downtime. What will a 22-hour power outage cost your business?
- N+1 power at the UPS is sufficient. Actually, just having an additional UPS and no other redundancy is only marginally better than no redundancy at all. UPS failure is not very common; much more likely is a failure somewhere between the UPS and the devices.
While no single solution will be right for everyone, there are some common steps that all IT managers can take to mitigate risk and to prepare for cost-effective implementation of a power redundancy plan.
- Conduct a rigorous assessment of what your power redundancy requirements actually are. Base your power redundancy budget on the estimated costs (including opportunity costs) of expected downtime.
- Not every device needs the same level of redundancy.
- Don’t buy more reliability than you need.
- Group non-critical systems in a non-redundant power network.
- Consider backup generators or a second power feed from the utility company. In-house redundancy does not mitigate the costs of failure from outside.
- Don’t wait for your power devices to fail. Develop and implement a preventive maintenance and replacement plan for your power system. This also provides predictable budget and spending, which your CFO will appreciate.
Hopefully, now that you’re armed with this knowledge about power networks and redundancy, you can start effective planning to keep your systems up and running. AIS can help you decide on and implement the power redundancy plan that is right for you.
About the Author
Ken Carter, vice president of data center operations and critical infrastructure at AIS, is a professionally certified Data Center Design electro-mechanical engineer. Carter has over 15 years experience designing, building, and maintaining Tier II-III data centers all over the country for companies such as LexisNexis and Fidelity Information Services. Meeting client expectations of ensuring continuous reliability, availability, and productivity is the forte’ of uptime and has been a key focus for Carter.