So Twitter put out a number this morning that I can't stop thinking about: 327,452 tweets per minute at the peak of last night's election coverage, right around the moment the networks called it for Obama. That's the new record, beating the old mark set back at the Olympics closing ceremony this summer. Three hundred and twenty seven thousand tweets. In sixty seconds. I don't think I can even type that fast, let alone have a coherent thought worth sharing.
I was doing the thing everyone I know was doing last night, which is watching one screen (the actual election results on TV) while holding a second screen (my phone, open to Twitter) and basically ignoring the TV entirely because the feed moved faster than any anchor could talk. Around 11pm my timeline turned into pure noise, the good kind of noise, everyone shouting the same three words in different fonts. My replies tab was unusable for about twenty minutes. I gave up trying to read anything and just watched the number of new tweets ticking up in the little "X new Tweets" banner, which itself started lagging because apparently even the client-side polling couldn't keep pace.
What I actually find interesting here isn't the politics of it (this isn't that blog and I'm not going there tonight), it's the infrastructure question. Somewhere there are engineers at Twitter who spent the last several months load-testing for exactly this night, and it mostly held. Compare that to four years ago, when the site face-planted repeatedly under way less traffic. There's a real story in how fast "handle a spike" went from "we'll probably be fine" to "we built for this specific night." Facebook apparently had similar numbers on their end too, something like 71 million people generating over 500 million interactions related to the election over the course of the day, though I trust that stat less since Facebook's counting methodology has always felt a little generous to itself.
The other thing nobody talks about is how bad the signal-to-noise ratio gets during something like this. My feed was maybe 40% actual information, 40% jokes (some genuinely funny, most not), and 20% people announcing they voted, which, fine, but did you need three separate photos of your sticker? I unfollowed nobody, because I'm not a monster, but I definitely muted a couple of people in my head for the night.
Small tangent, because I can't help it: I finally picked up an iPad mini on launch weekend, the $329 model, and I ended up doing most of my scrolling last night on that instead of my phone. It's a genuinely nice size for one-handed reading in bed, better than I expected honestly, though the display is noticeably worse than the regular iPad's retina screen once you've been staring at Twitter text for two hours straight. I keep going back and forth on whether I should've just gotten a Nexus 7 for $130 less and called it a day. The screen quality gap bugs me more than it probably should for a device I'm mostly using to read tweets in bed.
Back to the traffic numbers for a second, because there's a broader pattern here that I think is underrated: elections, award shows, sports finals, these are turning into the load tests that actually matter for consumer web infrastructure, more than any planned Black Friday traffic spike. You can schedule for Black Friday. You can't schedule for the exact minute a race gets called, and that minute is when everyone shows up at once. Whoever's on call at these companies tonight earned whatever they're getting paid, doubly so if they work anywhere near the networks' own sites, which from what I saw were also getting hammered pretty badly around the same time.
Anyway. I refreshed my timeline probably four hundred times last night for no good reason, same as everyone else, and this morning my eyes feel like sandpaper. Worth it, mostly for the jokes.