According to one report YouTube has 1.3 billion users, 5 billion downloads a day, 3.2 billion downloaded hours a month while 300 hours are uploaded to YouTube each minute.
Content Author was: Statistic Brain
Date research was conducted: September 1, 2016
Just some thoughts, happy to be corrected
Google currently has about 2.5 million servers with YouTube downloads now providing a sizable part of their advertising revenue
Calculations re YouTube Downloads
According to those stats in *September 2016, users download 5 billion videos a day, totalling 107 million hours (that's 6,420 million minutes). That's 3½ million videos a minute, totalling 4½ million minutes.
That means one video download is, on average, 77 seconds, some much more, some get interrupted, some are less.
At an average download speed of, say, 30-40 mbps, each 77 second video has an average resolution, a size of, say, 5-10 megabytes (that's 40-80 megabits) and would take about 1-2 seconds to download.
During that streaming minute a YouTube server could be looking after heaps of other clients, say 100 clients each minute.
So the 3½ million clients every minute could then, perhaps, be serviced by 35,000 high speed servers.
*According to a Variety news report, in late 2016 it reached 1 billion hours a day, that's ten times higher, not saying they reach that every day, but yes, they would need to scale up 10 to 20 times with more high speed servers. I'm assuming servers are monitoring that traffic many times per second.
5 hours (18,000 seconds) of download viewing time get uploaded every second.
Let's say that involves about 10,000 clients each second, and if each one could get its own dedicated YouTube server, no problems.
Each server should probably be able to handle 5-10 multiple clients, however.
In the third quarter of 2017, Netflix reported 109.25 million streaming subscribers worldwide. Of these subscribers, 52.77 million were from the United States.
Each Netflix account can feature a maximum of five separate profiles, enabling a better, individualized recommendation system.
To provide this service, it currently uses tens of thousands of Amazon Web Servers
** End of Page