This article is also available as a PDF download.
Hurricanes, tornadoes, and earthquakes aren't the only disasters that damage PCs, servers, and other computers. The abrupt loss of electricity prevents systems from closing open applications, completing replication actions, finishing disk activities, and shutting down properly. Lost data and corrupted databases and applications often result.
Powerful electrical spikes also cause trouble. A computer's sensitive electronics can easily be destroyed by electrical surges spawned by lightning strikes or power grid fluctuations.
Uninterruptible power supplies (UPSes) help prevent the damage that occurs from both power loss and common electrical fluctuations. However, just plugging in a UPS and connecting computer equipment doesn't guarantee systems are properly protected. Keep these 10 items in mind to maximize the protection UPSes provide.
#1: Data line protection is critical
Computers connected to UPS devices still aren't protected from the ever-present risk of lightning strikes. Sure, the UPS may guard against the PC or server being damaged by electrical surges reaching the system via a power outlet, but lightning strikes just as easily travel telecommunications links, such as cable modem and DSL connections, to toast everything in their path (including motherboards and attached hardware).
Ensure that you configure your network properly. Install a UPS data protection circuit between your telecommunications provider's data line and your computer systems. Typically, the telecommunication circuit plugs into a physical input on the UPS, which also presents an equivalent RJ-11 or RJ-45 outlet for continuing the circuit's connection to your local area network.
Should a strong electrical surge occur, the UPS' protective circuit will prevent the surge from destroying other equipment. The circuit may no longer function following such a strike, but replacing a UPS device is much less expensive than replacing multiple servers, PCs, and network switches.
#2: Recovery software must be installed
Often, it's tempting to just plug in a UPS, connect a computer or server, and get to work. Unfortunately, UPSes usually include proprietary software that must be installed for the unit to step in and properly power connected systems when the principal power source fails.
Always install a UPS' included software. In addition, be sure to connect the communications cable (typically USB or RS-232) to ensure that the PC or server and UPS can communicate.
#3: Recovery software must be configured
In addition to installing the proprietary software included with a UPS, you're best served configuring the program to meet your organization's specific needs. Although you can accept the manufacturer's default configuration, out-of-the-box settings may not allot sufficient time for shutting down a domain controller that must reconcile database, Active Directory, e-mail server, and other connections. Be sure that a system receives sufficient time to close all open connections and applications and properly shut down. Configuration options vary by manufacturer and model. In Figure A, an APC SOHO device's Power Chute software enables specifying shutdown settings.
Review the default settings to determine whether they require adjustment. Often two options are available. You can specify either how long a system will run on batteries before initiating a shutdown sequence or you can specify that the UPS begin shutting down attached systems when the its battery provides only enough power to operate for a certain number of minutes.
When configuring these settings, note that in both cases, the trigger focuses on initiating shutdowns based on battery values (either how long the battery's been running or how many minutes the battery can continue powering the current load). Regardless of which option you select, allow sufficient time for the shutdown sequence to complete. Some servers may require six, eight, or even 10 minutes to properly shut down; if the shutdown sequence is initiated when the battery possesses only five more minutes of power, data loss could result.
#4: Batteries fail
Batteries fail. It's a fact of life. Due to a completed lifecycle, manufacturer's defect, or some other cause, there's no guarantee a battery will operate as intended. Thus, even a high quality UPS doesn't eliminate the need for ensuring data backups routinely complete properly.Fortunately, UPS batteries typically provide warning signs that something's amiss. An online model may not properly power load levels, offline models may not test well, and still others may light warning or failure indicators. Regularly testing a battery (Figure B) helps ensure that it has sufficient capacity to power a system should the principal electricity source fail.
Regularly test a battery to make sure that it retains sufficient strength to power systems properly should the site's electrical supply fail. With systems supporting automated testing, schedule routine quick and deep cycle tests. Review the test logs regularly and keep an eye out for any anomalies that could indicate that failure is imminent. Always replace troubled batteries before they fail, as data loss could result if a failing battery remains in service.
#5: Load levels change over time
Think how your network has grown over time. New servers, workstations, and peripherals all place greater demand on UPS devices, but in most cases, the UPS was likely deployed with the original network—when power requirements were less.
Following expansion, be sure to recalculate the volts-amperes/wattage a UPS must provide. Whenever a server rack or workstation's equipment exceeds a UPS' capacity, either deploy an additional UPS or purchase a new model.
# 6: A UPS shouldn't power a laser printer
Printing may prove a critical task, such as within billing or credit departments, but don't rely upon a UPS to power a laser printer. Even if the UPS is being used only to condition the electrical supply a laser printer receives, when attached to battery backup outlets on the UPS, laser printers quickly overpower a UPS and exceed standard load levels. The quickest way to generate multiple load level warnings and alerts (and potentially damage UPS electronics) is to power a laser printer using a UPS.
Why? A laser printer's fuser (which melts printer toner applied to paper) consumes a quick burst of energy. Most UPS models simply can't supply the required power as quickly as the laser printer demands.
#7: Network connectivity is key in planning emergency power requirements
When powering servers or even a workstation that serves as a workgroup server, it may be necessary to complete data replication, communications, or other activities requiring network connectivity when shutting down systems due to electrical outages. Thus, any network equipment between the systems performing the replication or data sharing must also receive power from UPS devices. If network switches, routers, firewalls, and any required telecommunications modems or routers don't also receive power from a UPS during an outage, the data replication and communications links will fail (and data loss could result).
#8: Backups must terminate
This is often a forgotten setting: Be sure to instruct Windows Backup (and third-party backup routines) to abandon backup operations if battery mode begins. Also, when using Windows' native backup utility, specify that the task not start if the system is running on batteries. Select the battery-related check boxes within Windows Backup's Power Management settings (Figure C) so that incomplete backups don't overwrite properly completed backups.
These settings are not Windows' default configuration. Unless a site experiences frequent power outages lasting just a few moments, consider entering the changes to prevent an incomplete or corrupt backup from overwriting a proper, previously completed backup file.
#9: Service life is short
The service life of a typical UPS battery is only a few years. UPS devices are often positioned on cubicle floors and behind desks. It's easy to forget they're there. Years can pass quickly.
As we mentioned in item number four, batteries will eventually begin to fail. After a few years, batteries may continue to function but they likely won't continue providing the same levels of service. For example, a two-year-old battery may provide only 12 minutes of power versus 18 when it was new. Be sure to consider a battery's age and associated service degradation when reviewing your emergency power requirements.
Many UPS devices feature hot-swappable batteries. Such models enable changing a UPS' batteries without powering down attached equipment. Hot-swappable batteries are particularly useful when powering servers and other critical devices.
Thunderstorms, electrical sags and surges, and other conditions (including heat) can also shorten a UPS' service life. Ensure that a UPS receives a fighting chance at longevity by unplugging unused systems during storms, regularly checking battery strength, restricting load levels within reasonable limits, and keeping UPS products free of clutter (thereby giving them room to breathe and cool). If you do choose to unplug an unused UPS during thunderstorms, make sure that it's unused. In other words, verify that it's not protecting a telecommunications data circuit through which a lightning strike could travel and destroy connected equipment.
#10: UPSes are lifeboats, not bridges
When deploying UPS devices, never consider them bridges over troubled waters. Remember that UPSes are not generators. They're best used as lifeboats for escaping unexpected crises.
Although a UPS can certainly provide full functionality during a five-minute blackout, the devices shouldn't be used to bridge periods when electricity fails. Instead, if you view UPSes as a temporary solution for properly closing applications and processes and shutting systems down in an orderly manner, you're much less likely to suffer data loss, corrupted files and applications, and other failures as the result of blackouts.