Ask HN: How would you host video if you were building “the next TikTok”?
Say you are a startup with minimal funding, so costs count for alot.
How would you host the video? There’s no chance of doing it cost effectively with the major clouds because they charge 9 cents per gig…. you’d be broke instantly.
So what are some other cost effective ways to do the hosting?
8 comments
[ 2.7 ms ] story [ 23.7 ms ] threadOr it sounds like you might be willing to sacrifice hosting quality and do something like peer to peer storage and serving but that would be a terrible experience for user. Videos would often just not play.
I think you need to raise money for this startup and pay a major cloud provider?
No, I’m asking what the most cost effective ways are to host a lot of video serving.
There are other alternatives to big clouds. In fact you’d be crazy to serve video from big clouds even if you had the funding…. 9 cents or more per gigabyte is a huge amount of money.
Also, probably a dumb idea: can you use cloudflare images for gifs?
Maybe you can only post 1 second videos, maybe you can only post one video a day, or you only see the most important 10 videos of your friend. With some constraint, there could be alternatives.
https://lbry.com/faq/embed
That might get you enough headroom to bootstrap revenue to hosting your own servers colo.
The problem here would be defining the constraints. TikTok like video does mean you aren't going for long form and it's not a huge deal to do UHD or whatever. So you figure out the bounds of the transfer and scale and then do the math to figure out how to take the singular viewer thing and layer in seeding three(?) videos partially for every single viewer.
So a viewer watching a video only needs to watch once and repeat but they are also given chunks of two other videos to stream. I'd look at flocking algorithms maybe to test out how to design something like this. But I'm also an idiot.
I think there's something to that. I might try to make a proof for it at some point. I... do not want to run a social media platform or work on one. That sounds like a nightmare, but I think something like that might be not just possible but inevitable.
You could monetize it by having people pay for seeds and storage for their videos. You could make a marketplace that pays people to seed videos and they could choose what videos to seed so they don't seed stuff they don't care for. I dunno, it almost sounds like a media sharing protocol that should already exist. Does this already exist?
A rough proxy as described is a few hundred lines of Python.
https://news.ycombinator.com/item?id=28682458