From the Lab

Data Center Outages: High Stakes and Multiple Causes



Outages happen. Whether a data center outage is a nuisance or a disaster depends on how long it lasts and how critical those operations are to a business or organization.

Very short disruptions could be crippling. Losing access to high-performance computing could lead a financial firm to big losses in a matter of seconds. The absence of electronic health records at the moment they are needed could prove fatal in an emergency medical scenario. In other industries, such as e-commerce, losses grow significant over time.

Unexpected outages are never welcome, but some may be anticipated. The traditional (if not mythical) gold standard for telecom carriers is “five nines,” or 99.999 percent availability. For data centers, which rely on carriers, that number is lower. Amazon typically offers 99.95 percent availability on its cloud platform, which translates to about five and a half hours of downtime per year.

The causes of outages are diverse. A Data Center Knowledge article last year listed 10 of the “strangest” data center outages. It includes the following: an atomic clock leap-second glitch; squirrels; a data center migration; undersea cables impacted by ship anchors; stolen networking cards; a smoke alarm system responding to a smoldering cigarette; a power transformer that exploded upon impact with a truck; an errant border gateway protocol (BGP) message; theft (again); and Superstorm Sandy.

By defying the usual pattern of tropical storms, Sandy was freakishly, if not uniquely, destructive. (Katrina surpassed Sandy in assessed financial damage.) As a category, extreme weather drives most power outages, which in turn disrupt many data centers, and the economy at large.

A report prepared two years ago by the White House and Department of Energy named “severe weather” as the leading cause of power outages. It also said that aging electrical-grid components were a contributing factor. The grid has other vulnerabilities. Congressional hearings in the summer of 2015 drew attention to threats posed by solar storms and electro-magnetic pulses from nuclear explosions.

Solar and nuclear radiation pose special challenges, but a data center can nonetheless brace itself against a wide range of threats, from squirrels to a loss of power and beyond. Keystone NAP is a case in point.

Located on a site safe from urban hazards but convenient to several major cities, Keystone NAP features nine rings of physical security; offers a diversity of carrier-neutral connections, with multiple physical network paths leading to and from the site; and leans on a power infrastructure of unprecedented durability that aggregates more than 2,040MW from four industrial-grade power feeds within a few mile radius.

At Keystone NAP, we also realize that what you ultimately need is not just an array of available servers, but also an optimized combination of software and hardware. Thus we offer advanced “up the stack” services for business application and support to help you take full advantage of our robust facility.