Incase 1 of the clouds fs up and goes fully down

Ew kubernetes do they still use docker

But yea cloud kubernetes would prob be the alternative

To avoid dealing with instances

Spread across aws and gcp

Incase 1 goes down

You realize there is still a (small) chance that Cloudflare goes down, right?

So even if you had anything on CF's edge, you'd still have a single point-of-failure, just the same as if you relied on AWS.

Yea but its easier to deal with cloudflare

You can prob get like 50k open websocket connections on a $5/month DigitalOcean droplet 😉 I get you want to use DO's .. but it doesn't sound very economical for your use case (at the moment)

I'm think if we hit the scale we are aiming for with DOs maybe we could talk to billing for usecase where we just need 24/7 connected WS to reevaluate the billing for walltime

If we dont well it wont cost much eaither

Like netflix running on AWS is physically impossible with bandwith costs but they made a deal

Yeah, for sure, reach out to Sales, and they'll try their damndest to get something to work for you.

But it does sound like you'd be pushing up against the limits of what's possible without that extra billing/engineering support.

i think a lot of people are going to have similiar problems with websockets and wall time

Yea but do you want to deal with instances? Its alot of responsibily as soon as you need to "spin something up" you open a large rabbit hole. (Kernel updates, baremetal maintenance, baremetal network issues, security updates, legal issues due to not implementing security update ontime, userspace segfaults, kernel segfaults, instance degradation due to overprivisoning the host, instance dosing, i could go on)

Honestly that sounds exactly like what Kubernetes tries to solve haha

terraform and docker will fix this problem

There is a cost attached to fully managed solutions

Yea kubes would be option 2, but we need alot more software coded for that to work

in effect youre paying cloudflare or whoever the extra money to do all that maintenance for you - its a hassle/cost tradeoff that only you can choose

More software, more bugs, longer development cycle, slower to market time

The to market time is priceless

What if by being in the market 3 months earlier you actually added compound usergrowth lasting 3 years

well its up to you where u want to find the balance. cost or development time

That could be worth 100s of millions of dollars

No way to scientifically measure this stuff tho

If you're seriously chatting about 100s of millions of dollars, you definitely need to reach out to Sales.

or to me... I'll do it for only one of those hundreds