I’m not here to shoot down your comment in response, but you are confusing a few different things here. My point was just to demonstrate how the actual cost for streaming from a provider when you throw all the extra junk they do on top of it. You can have very reliable, distributed systems without all that mess, they just choose not to go that route. Selling a user subscription is just the beginning down a rabbit hole of upsells, price hikes, content lures, marketing gimmicks, data capture and sale programs, Ads (for a service you already pay for in most cases)…it’s expansive. Each one of those things has a team behind it making decent money.
If they just wanted to stream things they could with much less cost and effort, AND make money doing it. It’s been done before. They all just choose to go the route of squeezing their audience for every last ounce of possible monetization they can, which costs a ton of money.
Do you really think that k8s, which is free (unless you buy a support contract) is less easy to manage at that scale than, what - a bunch of VMs with Docker installed? PHP and a NAS? Is that REALLY the cost sink you think it is?
Brother, what works in your homelab will NOT survive Netflix traffic levels.
And streamers are raising prices to get rich. Not because “k8s hard” lol
Hey, friend. No need to get mad. I’ve worked on these exact systems I’m describing for years, which is how I know these things. They suck. It’s not about k8s itself, you’re focused on the wrong part of what I’m saying.
You want to just stream files to people cheaply and quickly, these companies don’t do that anymore, and that’s the choice these companies have made. They’d rather scale their engineering resources to the microservices that nickle and dime the shit out of everyone. That’s just the facts.
Ok but marketing teams and micro transactions teams to get rich have fuck all to do with k8s or clusters to run lots of data quickly over the network.
Maybe you know some magic way to provide high availability for petabytes of data pushed at hundreds of Gigabits per second to millions of simultaneous connections. But I don’t. And I work with large data volumes for research that still are a fraction of Netflix data and throughput and we’ve hit the limit of single high performance storage devices necessitating a ceph cluster for storage and availability.
I’m not here to shoot down your comment in response, but you are confusing a few different things here. My point was just to demonstrate how the actual cost for streaming from a provider when you throw all the extra junk they do on top of it. You can have very reliable, distributed systems without all that mess, they just choose not to go that route. Selling a user subscription is just the beginning down a rabbit hole of upsells, price hikes, content lures, marketing gimmicks, data capture and sale programs, Ads (for a service you already pay for in most cases)…it’s expansive. Each one of those things has a team behind it making decent money.
If they just wanted to stream things they could with much less cost and effort, AND make money doing it. It’s been done before. They all just choose to go the route of squeezing their audience for every last ounce of possible monetization they can, which costs a ton of money.
Do you really think that k8s, which is free (unless you buy a support contract) is less easy to manage at that scale than, what - a bunch of VMs with Docker installed? PHP and a NAS? Is that REALLY the cost sink you think it is?
Brother, what works in your homelab will NOT survive Netflix traffic levels.
And streamers are raising prices to get rich. Not because “k8s hard” lol
Hey, friend. No need to get mad. I’ve worked on these exact systems I’m describing for years, which is how I know these things. They suck. It’s not about k8s itself, you’re focused on the wrong part of what I’m saying.
You want to just stream files to people cheaply and quickly, these companies don’t do that anymore, and that’s the choice these companies have made. They’d rather scale their engineering resources to the microservices that nickle and dime the shit out of everyone. That’s just the facts.
Ok but marketing teams and micro transactions teams to get rich have fuck all to do with k8s or clusters to run lots of data quickly over the network.
Maybe you know some magic way to provide high availability for petabytes of data pushed at hundreds of Gigabits per second to millions of simultaneous connections. But I don’t. And I work with large data volumes for research that still are a fraction of Netflix data and throughput and we’ve hit the limit of single high performance storage devices necessitating a ceph cluster for storage and availability.
You’re just making my point.