Meanwhile I moved one of my two internal DNS servers to a second site on 11 Nov 2020, and it's been up since then. One of my monitoring machines has been filling, rotating and deleting logs for 1,712 days with a load average in the c. 40 range for that whole time, just works.
If only there was a way to run stuff with an uptime of 364 days a year without using the cloud /s
> the point is that when it's down, bring it back up is someone else's problem.
When it's down, it's my problem, and I can't do anything about it other than explain why I have no idea the system is broken and can't do anything about it.
"Why is my dohicky down? When will it be back?"
"Because it's raining, no idea"
May be accurate, it's also of no use.
But yes, Opex vs Capex, of course that's why you can lease your servers. It's far easier to spend company money with another $500 a month on AWS than spend $500 a year for a new machine.
My lightswitch is used twice a day, yet it works every time. In the old days it would occasionally break (bulb goes), I would be empowered to fix it myself (change the bulb).
In the cloud you're at the mercy of someone who doesn't even know you exist to fix it, without the protections that say an electric company has with supplying domestic users.
This thread has people unable to turn their lights on[0], it's hilarious how people tie their stuff to dependencies that aren't needed, with a history of constant failure.
If you want to host millions of people, then presumably your infrastructure can cope with the loss of a single AZ (and ideally the loss of Amazon as a whole). The vast majority of people will be far better off without their critical infrastructure going down in the middle of the day in the busiest sales season going.
Many B2B-type applications have a lot of usage during the workday and minimal usage outside of it. No reason to keep all that capacity running 24/7 when you only need most of it for ~8 hours per weekday. The cloud is perfect for that use case.
Scaling itself costs nothing, but saves money because you're not paying for unused capacity.
The main application I run operates in 7 countries globally, but the US is the only one that has enough usage to require additional capacity during the workday. So out of 720 hours in a 30 day month, cloud scaling allows me to pay for additional capacity for only the (roughly) 160 hours that it's actually needed. It's a significant cost saver.
And because the scaling is based on actual metrics, it won't scale up on a holiday when nobody is using the application. More cost savings.
Nice of you to assume that I don't understand the pricing of the services I use. I can assure you that I do, and I can also assure you that there is no such thing as provisioned vs on-demand pricing for Azure App Service until you get into the higher tiers. And even in those higher tiers, it's cheaper for me to use on-demand capacity.
Obviously what I'm saying will not apply to all use cases, but I'm only talking about mine.
