The bizarre trigger that led to Twitter’s tweets-per-second record being obliterated earlier this month provides a powerful reminder to IT departments everywhere: You never know when or why your traffic will spike.
In Twitter’s case, it was a Japanese broadcast of anime master Hayao Miyazaki’s beloved "Castle in the Sky." Not a huge sporting event, juicy celebrity revelation or political scandal, but rather hundreds of thousands of Japanese anime fans feverishly posting the word “balse” all at once. (The word is uttered during the film’s climactic scene; the exact same phenomenon set a previous record for Twitter in 2011.)
The recent resulting flood of tweets peaked at 143,199 tweets per second, or 25 times the social network’s typical flow. That shattered the former record of 33,388 tweets per second set last New Year’s Eve.
More importantly, it showcased Twitter’s re-architected infrastructure, which in years past would have crumbled under the onslaught. In fact, in a blog post detailing the record-setting event, Raffi Krikorian, Twitter’s VP of platform engineering, clearly took pride in Twitter’s ability to withstand the barrage.
“During this spike, our users didn’t experience a blip,” Krikorian wrote. “This goal felt unattainable three years ago," he added, referring to the 2010 World Cup, when repeated waves of tweets caused Twitter fits.
Airings of “Castle in the Sky” are hardly the only triggers of such unexpected traffic spikes. For instance, when reports of the National Security Agency’s spying activities were leaked--implicating Google and other large tech companies for providing the NSA with data--traffic surged on a number of alternative search engine sites, all of which appeared to effectively handle the increase in visitors.
[Read how this year's Super Bowl web traffic crushed sites from Coca-Cola and other brands that blended interactive marketing with televised commercials in "Super Bowl Web Traffic Blitzes Advertisers."]
The NAACP, however, didn’t fare so well when it circulated a petition last month calling for federal prosecution of George Zimmerman in the Trayvon Martin case. The petition had 350,000 signatures within hours, but the resulting traffic caused the site to crash.
Even when companies expect traffic spikes--such as during the holiday shopping season--problems can occur if companies aren't fully prepared. Just ask Best Buy, which had to actually reduce the flow of traffic into its site for an hour on Nov. 19, 2012, when it was overwhelmed by a surprising number of early holiday shoppers. Or Sears, which somehow was completely unprepared for the amount of traffic that flooded its website on Black Friday in 2008, forcing the company to shut down the site for two hours on the biggest shopping day of the year.
Gartner analyst Colin Fletcher said via email that events such as Twitter’s anime surge highlight the need for organizations to ensure they have the technologies in place to prevent any unnecessary outages.
“More than ever before, you’ve got to double down on network configuration management and performance monitoring,” Fletcher said. “Network configuration management is still criminally underutilized as a way to quickly and confidently make critical changes to network services on demand, and most companies are utilizing monitoring tools and techniques that can easily miss crippling ‘bursty’ events.”
While they’re auditing the tools at their disposal, perhaps some network admins should also brush up on a few fundamentals of website preparation because the next traffic surge to besiege their sites may come any second, for whatever reason.