Cascading failure

There’s a term for when a single hiccup triggers a chain reaction that makes everything go absolutely, altogether, totally, and undeniably wrong, causing a large and intricate system to collapse on itself. On the street you might call it a fiasco—but in more formal parlance it’s called “cascading failure.”

Sound familiar? If you’ve been through an electrical grid outage, there’s a good chance you’ve heard it in that context. It’s not a new phenomenon, but it’s a relatively recent term, and the complexity of modern life has multiplied the real-life scenarios for its use in the fields of technology, biology, and finance. The easiest way to think about cascading failure is as a line of tumbling dominoes—or the plot of Jurassic Park, a blockbuster about how the smallest of errors can lead to total catastrophe.

1: Mistakes or people required to start a chain reaction

2-2.5: People to which each person infected with the coronavirus is expected to transmit the virus, causing it to spread exponentially

20: Minutes it took for the stock market’s “flash crash” in 2010, part of a cascading failure related to high-frequency traders

11%: Share of transmission lines in the US-South Canada power grid that are “vulnerable to the kind of ‘primary failure’ that could trigger a cascade”

2%: Share of US-South Canada transmission lines that cause 85% of cascading failures

670 million: People plunged into darkness in India in July 2012 after the world’s biggest power failure

10%: Share of the world’s population affected by the 2012 India blackout

The blockbuster film (and novel) Jurassic Park turns the sudden, terrifying loss of control represented by cascading failure into thrills. When the movie’s IT-employee villain shuts down a security system to hustle embryos out of the dinosaur preserve, it triggers a series of disasters that leads to dinos running wild and devouring the guests.

The film’s hero sees disaster coming because he’s a chaos theorist, a somewhat obscure branch of math given a star turn in the film. While it’s debatable whether the path that destroys the park is true chaos or mere extreme complexity, researchers have used chaos theory to investigate the technology most associated with cascading failure: the power grid. As Peter Fairley writes for IEEE Spectrum, one of the breakthroughs came in 1982 when mathematicians applied chaos theory to the subject and found its (paradoxically) familiar patterns everywhere.

While avarice is at the heart of the disaster in Jurassic Park, much of Italy lost power in 2003 because a single tree, compelled by gravity rather than greed, fell on a high-voltage power line. With fewer lines to support the electrical load, the system strained its capacity, and just 24 minutes later, a tree tripped a second line, perhaps “caused by the sag in the line, due to overheating of the conductors” (pdf). About three minutes later, the entire system collapsed. (One dilemma in designing failsafe power grids is that warnings of a failure cannot travel faster than the electricity itself.)

Meanwhile, power grids across the world have only gotten more complex, and not just because of increased demand. Renewables like wind and solar offer less predictable loads. Interconnecting power grids make systems more robust under normal circumstances, but add more failure points. The more technology supports us, the more stress it’s under.

“One ingredient of many fiascos is that great, massive, heart-wrenching chaos and failure are more likely to occur when great ambitions come into play.”

—Ira Glass, host of This American Life

1965: A massive power outage hits the northeast US during rush hour, plunging 30 million people into confusion, due to a single protective-relay failure; the term “cascading failure” begins to appear in US government documents.

1984: A wire-service translation error causes a bank run on Continental Illinois, leading to the then-largest bank failure in US history.

1988: Scientist Eric Drexler outlines an apocalyptic “gray goo” scenario, in which self-replicating nanotechnology accidentally runs wild and devours the planet.

1999: The then-largest power outage hits Brazil from March 11-June 22, and begins with a single lightning strike.

2010: In just 20 minutes, a single large sell order sends the New York Stock Exchange on its biggest free-fall in decades, a “flash crash” causing the Dow Jones index to lose almost 9% of its value.

2012: For still-unknown reasons, 670 million people in Northern India lose power in July 2012 after a chain reaction caused by “power trips” along the Bina-Gwalior electrical line—the largest chain-reaction outage to date.

2014: A Boeing Dreamliner suffers from a thermal runaway fire: “one [battery] cell going boom and taking out the next cell and the next cell and the next cell.”

2016: China breaks up part of its power grid to prevent cascading failures.

Heavy footsteps are what researchers believe caused the 6,600 photomultiplier tubes in the Japanese Super-Kamiokande neutrino detector to implode in November 2001. The facility contains a 50,000-ton water tank buried in a zinc mine under Mount Ikena in Japan. The tubes are light-detecting and can capture a radioactive glow produced when neutrinos smash into atomic particles in water, and was used in 1998 to prove that neutrinos—subatomic particles with no electrical charge—have mass.

About half of the 11,000 glass light-detector tubes lining the underground laboratory were destroyed in a chain reaction caused by routine maintenance workers (who were wearing protective foot gear) as they stepped on a couple tubes.

Space may look empty from here, but it’s full of junk orbiting the Earth at 22,000 miles per hour (35,400 km per hour)—at which a paint fleck carries the force of a motorcycle at highway speeds. In the 1970s, NASA scientist David Kessler predicted that once the debris around the earth reaches critical mass, one collision could set off a zero-gravity demolition derby and wipe out swaths of critical infrastructure, or worse yet, the International Space Station. “A 10-centimeter sphere of aluminum would be like 7 kilograms of TNT,” a NASA scientist told Wired. “It would blow everything to smithereens.” (The 2013 film Gravity is inspired by Kessler syndrome.)

Yet there are plans to add tens of thousands of new satellites, “constellations” that would surround the globe with reliable internet access. Each one would be at risk of a collision, and then causing collisions in turn. According to the director of NASA programs at Boeing, the collision-avoidance maneuvers that satellites have to take “could swell from a current average of three a day to about eight an hour,” reports Scientific American.

Watching cascading failure unfold is like watching a slow-moving car crash. It’s almost artful how a chain reaction can lead to total chaos, as seen in this video simulation of San Francisco losing power because of a malfunction at one power station.

Is the spread of the coronavirus a cascading failure? It depends on where you are. In many cases, it’s a nation’s response to the spread of a virus that determines just how cascading a failure can be. In the United States, the lack of a quick response and a few bad policy decisions months before the virus took hold have proven to exacerbate healthcare system weaknesses as the spread of the virus takes hold. Consider this chain of events:

April 2018—US president Donald Trump’s administration dismantles a National Security Council team that oversees responses to pandemics.

Jan. 31, 2020—US Health and Human Services head Alex Azar declares a public health emergency, which actually narrows the number of labs that can make tests for Covid-19.

Feb. 7, 2020—The US Centers for Disease Control begins to distribute tests for Covid-19, but they don’t work. By mid-February, the US is testing just 100 people a day.

Feb. 26, 2020—The CDC and the US Food and Drug Administration give state and local agencies permission to use a workaround on the flawed test.

Feb. 29, 2020—The FDA permits certain facilities to develop their own tests. A former FDA commissioner tells the New Yorker that the dissolution of the NSC pandemic team is partly to blame for the US having a “single point of failure” on testing.

March 17, 2020—25,000-50,000 Americans have been tested, but the director of the Harvard Global Health Institute tells the Miami Herald it should be testing four to six times that many people every day.

March 31, 2020—The White House predicts up to 240,000 Americans may die due to the virus, in spite of efforts to contain its spread.