Mike Lindell claims that his new social media platform–which he describes as sort of like a combination of Youtube and Twitter–can handle a billion members. I’m wondering how many people can be handled by a streaming media server. I know only a fraction of the members of a service would be using it at any one time, but around how many servers would you reasonably need to serve a billion users? Are we talking on the order of 10,000? 100,000?
That’s basically where Facebook was at a few years ago, if you take it to mean monthly active users (a common metric.) You’re talking like two dozen very large datacenters worth. 100,000s is the right order of magnitude. Obviously, Mike Lindell is confused about the potential capacity of his site. It is nearly impossible to setup something this large in this amount of time, especially with the current shortages.
Mike Lindell seems to be confused about a lot of things.
This slide presentation by a couple of developers from Ye olde Year 2017 says they were getting “6000 HD streams @3Mb/s and up to 80 Gbps throughput per server”. That appears to be for live streams, so the number of simultaneous users is sure to go up a lot for cached streams, especially if you are allowed to set up more than a single server, which is imperative for any kind of real site anyway.
Anyone on an elastic cloud computing platform could legitimately say they could scale to a billion users. Give AWS your credit card number, set your server farm to ‘auto-scale’, and you can have as much capacity as you can pay for. The network capacity would be a bigger issue than the server capacity (if you actually have the load).
I expect that what will happen is that Liddell’s credit card will be maxed long, long before he reaches even a million users. The site will bog down under the load; he will blame Jeff Bezos for not providing cheaper computing infrastructure, and he will announce that he’s creating a new better AWS designed with free speech in mind.
Reading text messages and streaming video are two vastly different use cases. Just looking at bandwidth, a user reading text can consume, I’d say, at most about 50 characters per second. On the other hand, a user viewing even an SD video stream is consuming something like 150,000 bytes per second, more than three orders of magnitude more bandwidth. CPU resources to serve text vs. serving a video stream similarly differ hugely. So it’s very hard to answer this question without more details about exactly what data the service will provide and how users will be using it.