The mission of our company is to build products and services to help system administrators reach 100% uptime for their networks. We all know that it becomes harder and harder to reach 100% the closer you come to your goal, it’s like trying to reach the speed of light.
One of the limitations that system administrators can’t do much about is the reliability (or failure rate) of the electrical power grids they have to rely on. In a data center it is pretty easy to get around this problem by installing uninterruptible power supplies (ups): using a generator you can run a data center off-the-grid for days.
But what about the on-site networks that you serve with that data-center, the people in the offices and factory buildings. Your ups won’t help. When the power grid fails, your computer-using colleagues will twiddle their thumbs until power is coming back again.
Before you even try to go from three-nines (99,9%) to four-nines (99,99%) with your data-center you should look at the network your are serving: Maybe the investments make more sense there?
How big is impact of power grid reliability on your network reliability?
The global reliability of the power from the outlets in your wall depends heavily on the region/country you live in. I expected to find that the more developed a country is, the more reliable their power networks are. And in fact the power networks of less developed countries have the highest failure rates. But even in the developed countries the differences are huge:
- In 2011, Germany set a record with a downtime of only 15.31 minutes (as measured by SAIDI—System Average Interruption Duration Index—a common measure of reliability used in both the U.S. and Europe).
- In contrast the United States had a downtime—not including planned interruptions or extreme weather events—of 240 minutes in 2007 (It has been estimated, that grid downtime in the United States costs the US economy around 150 billion dollars a year)
That’s 4 hours of downtime for the average electrical power customer in the U.S.. Or 0,05% of the year, which means that the average non-ups-covered network could not reach an uptime of more than 99,95% in one year. If these 4 hours of downtime are all happening in one month the downtime is 0,6% and your uptime goes down to 99,6%. Ouch.
I live in Germany and I couldn’t easily recall the last time when I experienced a power outage. Consequently it is not very common here to have a UPS for your desktop computer or smaller data room.
When I researched for this article I was really surprised by the situation in the U.S., which seems to get worse over time, too.
The United States is generally considered to have one of the industrial world’s most reliable electric power systems. However, when compared statistically to other nations, the U.S. grid does not necessarily meet those expectations.
from: US Congressional Research Service report on storm related blackouts August 2012
Here is data from other EU countries (source):
If your European country is not listed, please see here for a longer list.
The above (Fig 1.2a of Terrorism and the Electric Power Delivery System) shows a measure of total electric grid outage duration across a sample of countries in the world.
The developing countries in the sample range from a few hours (Argentina) to a few days (Columbia) of outage per year. One powerless day means your maximum potential uptime is 99,7% for the whole year.
Most other countries are listed here. For example in Bangladesh, the country with the most failures, you experience a power outage on 248 days of the year. So for them server uptime is not the most pressing problem.
What’s your experience regarding the effects of the power network on your daily work as system administrator? Please share them in the comments.
PS: If you like how we think about IT and networks, check out our software PRTG Network Monitor which monitors uptime, usage and performance of networks.
Dirk Paessler's Personal Blog 
In italy it’s common to have a UPS on everything.
LikeLike