How Gangnam Style Broke YouTube - Computerphile

When YouTube was created, they had no idea that a video might amass over 2 billion views. Dr Steve Bagley explains why Psy's Gangnam Style music video forced YouTube to update their system.

This video was filmed and edited by Sean Riley.

[produKtNZ]

Did anyone notice a fancy bit of audio editing at 6:13? 
Steve gave the wrong answer first, but listen closely to it and hear the edit for yourself.
You had ONE JOB Steve! :D

[makismakiavelis]

3:00 why does everyone thinks that in youtube "1 person = 1 view"? Doesn't youtube view counter counts views and not unique visits?

[Marcsine]

Fun fact: 2^31 - 1, the largest positive value you can have in a 32-bit signed integer, is a Mersenne prime! Before Gangnam Style, you could use your maximum view count to do RSA encryption, I guess... Unfortunately, the new 2^64 - 1 is a compostie number, and therefor an ordinary Mersenne *number*.

[MattHaleUK]

"Sevennnty Seven"

Very creative editing.

[xwaltranx]

when i first saw the topic i asked myself: why not unsigned?
and here is the answer. thanks.

[MZZenyl]

I wonder how many million year it'd take, with the (dropping) average amount of views Gangnamstyle gets per day, to reach the new limit of 9, 223, 372, 036, 854, 775, 807.

[aserta]

It's not actually 2 billion people watching the video, it's 2 billion hits. It's not even 2 billion different computers/phones/tablets etc.

If you check the number of people on YouTube (and the latest i believe is in April of 2014) the count is at 1 billion people.

And on the internet there's ~3 billion?? (a rough estimate done somewhere in July of 2014).

Just thought that was relevant.

None the less very interesting video in the way computers deal with numbers.

[natminame]

his voice always reminds me of professor snape

[carlb.]

2:14 "Every time we add an extra bit, we double the amount of information we can store"
Sorry to nitpick, but you double the number of possible combinations you can store. The information you can store simply increases by 1 bit.

[bradster]

This guy is one of the clearest and precise teachers I've ever listened to.

[skye]

What I dont understand is, when it first started to approach a billion views, why didn't they increase the size ahead of time? They should have been prepared by then. Hitting 1 billion views was a big event, it should have been upgraded around then.
You think they'd learn from ipv4 that 4 billion might seem like a lot (2 billion signed), but it goes quick. 4 extra bytes per video is beyond insignificant when it comes to the data youtube deals with.

Its about to be 2015 and barring a handful of mobile devices, every computer worth supporting has 64bit architecture and we have an abundance of storage space- there is absolutely no reason to be using int32 instead of 64 in general purpose computing. Compilers should start defaulting unspecified ints to int64, and let us specify int32 (long) when we need to use something smaller for specific optimization tasks, just as we specify shorts and bytes as needed today.

As a computer scientist, I personally hope people start modernizing for once. We have the hardware and we are wasting it. This is the data typing equivalent of still having to design webapps for ie6, or supporting windows xp.

Lets please make 2015 the age of int64.

[bqfilms]

why would they store negative numbers?

[ivannic]

6:30 Would it really be a random number? If we store -77 as an unsigned integer, it would be stored as a maximum number minus 77, which in this case will be 4'294'967'219?

[AiOinc]

Still doesn't make sense why they needed to have a signed integer for view count, even if they wanted to check between dates, say date A has X views and date B in the future from date A has Y views, it's impossible to un-view a video, so you'd never get a number below 0, and the other way around makes no sense to use.

[SkyrimHod]

I think a better reason for switching to 64 bit instead of just an unsigned 32 bit number is simply the fact that it would only be pushing the problem a little further down the road if they simply switched to unsigned 32 bit.  If that video already has 2 billion views, it has a pretty good chance of eventually reaching 4 billion as well and if that one doesn't, some other one almost certainly will.  There's no way they'd ever reach the limit on a 64 bit number, so might as well just fix it the right way the first time around.

[JBinero]

I use unsigned numbers all the time. I do not think there is any point in using a signed number when it cannot be negative anyway. All that'll help with is introduce bugs to the system.

[Hedgeeeeee]

Simple solution:

If (views > 2, 000, 000, 000) {
Return 2b;
}

[magottyk]

Now you've done it, I've finally watched it.

[luisdanielmesa]

I have a few problems with what it's being said here... first, it doesn't matter if Google was using 32 bit computers, they could have chosen 64-bit or any bitcount number... and second, I've never seen "buggy" unsigned integer numbers (at least not in serious programming languages)... a calculator that doesn't calculate is just a paperweight.

[ZerMemry]

in 2038 all 32bit unix systems will also encounter this problem.