Common knowledge across the IT world is that power problems are a major cause of data loss.

An important note to keep in mind when considering the power protection implications at your site: The nature of datacenter-based computing mitigates (somewhat) the data corruption issues discussed below.

Everything concerning power in the datacenter (your servers and Windows desktop sessions) is protected by fantastic power protection:

• Separate 'A' and 'B' power delivery
• 'A' and 'B' Industrial UPS Systems
• 'A' and 'B' Diesel Generators
• Regular Scheduled Testing of UPS and Generators
• Etc.

The power has never gone down in our datacenter. Clean, reliable power is the number one commodity sold by the datacenter.

As an example of how datacenter-based (cloud) computing affects power protection calculations at your location, consider a power incident affecting a workstation PC that causes a Remote Desktop session to be disconnected. If a connection to an RDP session is lost, the Windows session is left running in the datacenter, waiting for someone to reconnect. No data is lost or corrupted in the datacenter. Since no corruption happened in the datacenter, one might consider the implications (cost, maintenance) of protecting the remote workstations.

In a practical sense, this has modified our power protection recommendations for our customers of datacenter-based computing. Where we previously recommended 685VA (390W, AVR) smart UPS's for ALL PCs, we now recommend only 350VA (255W, AVR) UPS units for ThinPCs and other thin-client devices at only 40% of the cost.

Our power protection recommendations remain that all computing and communications gear be protected by a UPS. However, one must consider the intended use (load) in choosing the proper size (cost).

Notebook (portable) computers can also change your power protection calculations. Since they already have batteries, protecting portable devices may be simpler (surge protection only).

According to an oft-quoted study by Bell Labs, 87% of all power disturbances are brownouts or dips in voltage delivered by the power grid. Brownouts are the biggest obstacles faced by computers and communications gear and can cause truly difficult-to-troubleshoot (expensive) problems (like data corruption). Another study by IBM indicates that a typical computer may see more than 120 power problems per month.

The repercussions of power problems (like down-time) are often multiplied by the number of users on your network.

Surge suppressors, at best, provide only limited power protection. The majority of surge suppressors in the field are partially or completely dysfunctional. They wear out, but still pass power and surges. We generally only use surge suppressors as power distribution devices (behind a UPS) and we don't generally care if they are working or not, as long as they pass power reliably.

Surge suppressors do not provide automatic voltage regulation (AVR). They only clamp (limit) over-voltages. Surge suppressors generally clamp at 330V to 500V…meaning your gear will see (be affected by) those over-voltage incidents. Damage will occur over time. Better-quality UPS units will regulate voltage (AVR feature, 'buck and boost').

The sizing of your UPS (in volt-amps or watts) in relation to the load will determine your optimal UPS size (run-time).

• We generally use a 685VA (390W) smart UPS with automatic voltage regulation (AVR) for most PCs.
• We generally use dual 1500VA (900W) smart UPS units with AVR for servers with dual power supplies.
  • Two 1500VA UPS units can probably support two running servers.
  • On-battery run-time is dependent on the actual load.

Yes, power protection is an expensive 'drag' that requires ongoing maintenance and cost. But it is perilous to ignore.

UPS units use lead-acid battery technology, like a car, and they fail predictably.

UPS units must be replaced about every three years.

Testing existing UPS units in-place is not simple and failing UPS units can cause very difficult-to-troubleshoot problems. Generally, we mark UPS units when we install them with the date of installation. Then, we refer to that date when any question of UPS efficacy comes up. We have found new UPS units are generally quite reliable up to 2.5 years of age. After that point, we just replace them rather than spend time troubleshooting them…it's just more cost-effective.

To test a UPS unit in place, we will generally reboot a PC just to the POST (BIOS) screen, then pause the boot (usually just by hitting the PAUSE or DEL key). Then we yank the power cord so that the UPS cuts over to battery power. If the PC stays up (doesn't immediately power off), we check the run-time. If the UPS will keep the PC up for 5 or 10 minutes, we call that good. While this is an unscientific test, it's fairly easy to do and indicative of the state of the UPS unit.

If the power goes out, and your PC continues to run on the UPS battery (beeping), shut down the PC gracefully so as not to damage the disk filesystem when the UPS battery runs out.

A few references (among a great many):

http://www.cbldatarecovery.com/blog/data-recovery/top-causes-of-data-loss

http://www.databackuponlinestorage.com/7_Causes_of_Data_Loss

https://www.imobie.com/support/top-5-causes-of-data-loss.htm

http://www.nerd4rent.com/reports/powerproblems.htm

http://members.tripod.com/thecomputerguy__/power.htm

http://www.techrepublic.com/article/10-things-you-should-know-about-deploying-a-ups/