It is easy to setup fault tolerant, simple, low maintenance and cheap for small traffic and data infra in the cloud (e.g. AppEngine+CloudSQL) where provider will provide 99.99% availability.
Internet not working is usually can be mitigated through multiple providers, cellular included
yes, and then you need to have engineer with $300k salary to support your app/postgresql cluster, load balancer, fallack, backups, security, etc with unknown quality.
Er. We did that years ago and it was perfectly normal. Some sort of load balancer appliance (I honestly kinda liked the F5s) on the front, your choice of compute cluster behind it running at least 2 copies of each service (usually on 2 separate hosts). I could be in a bubble but I was always under the impression that that was common, and it never seemed terribly hard.
I do find it fascinating that we are entering a time now where there are engineers who have _only_ ever used cloud services - and that they can't conceive of a era where _we used to have to do all that stuff ourselves on-prem_.
BTW this isn't a criticism - just an observation - cloud has been around now for that long. Still feels "new" to me :D
I worked in the shop which had all of these ops (app cluster, DB cluster) on premise, and they had 6 high paid people serving this, not just one, and I think there are many companies with their own sites like that.
> So you are saying when I am able to configure postgres I can earn a $300K salary? Sorry but this is utter bullshit.
did you ever try to configure postgres cluster with failover, and also backup, recovery, monitoring?
Also, I said it is hardest part, but you still need to know your load balancer, security, monitoring, updates, app deployment and configuration stuff.
Yes, this is big chunk of knowledge which in my understanding you don't have, and the topic is that this all worry free replaceable by AppEngine+CloudSql for $100/m if your traffic/data are not very large.
> yes, and then you need to have engineer with $300k salary to support your app/postgresql cluster, load balancer, fallack, backups, security, etc with unknown quality.
So? Cloud engineers are paid $300k/a to know all the 100s of little intricacies needed with putting your production compute nodes into a cloud.
You aren't simply going to be able to move your shit to cloud-only services and fire all the network/server/infra people without replacing them with cloud people.
Whether you're hosting locally, on the cloud, or somewhere in between, you need to pay a salary to someone to make sure that everything stays up.
The trade-off is that workloads and type of work differ between businesses: on one extreme a single technical person can move everything to cloud services and manage that f/time and on the other extreme a single technical person can move everything to local and manage that f/time.
Where the business is on that spectrum dictates what they should be doing to get the most bang for their buck.
Either way, no matter where you are on that spectrum, you are going to have at least a single technical person on staff. Asking for that crew to be cloud experts in addition to local-infra experts is going to be more expensive than hiring local-infra experts.
The TLDR is: unless you have actual concrete evidence that moving to (or starting with) a cloud solution is cheaper over a single financial year, it probably isn't.
> You aren't simply going to be able to move your shit to cloud-only services and fire all the network/server/infra people without replacing them with cloud people.
the point is that for simple app+db you can: AppEngine+CloudSql will provide full automation on few button clicks. You probably will need one "IT guy".
Internet not working is usually can be mitigated through multiple providers, cellular included