As far as I understand, many of the data centers are actually not owned by the big cloud providers. DC specific companies lease the space to the cloud providers and provide the staff to rack and stack. So, there is actually quite a bit of money in the data centers themselves.Interesting. AI powered data centres are consuming as much powered as a medium sized town.
But the AI data centres aren't making money (this is mainly a scramble amongst Facebook/MS/Google to acquire a monopoly on users)
& the AI results aren't correct (wrong summaries of youtube videos for ex.) Or useful (Office 365 with AI had lesser value than regular office)
Everything is monetized because everything has a cost. Hardware, software, licensing, rack space, bandwidth, power, man hours.Yea these data centers are big business. But I assure you they are multi use. Or just for AI. And if they are contracting for power feeds chances are these are MS owned. There are a plethora or colo centers out there with business specific areas that are literally gated or walled off from each other. Just cheaper that way. In reality cloud services are eating these places unless they just sublet that space back to cloud services.
And don't get me started on cloud services good Lord they charge for everything at the enterprise level.
You want networking... Charge
You want to store data.. charge.
You want to move data.. charge
You want to delete data.. charge.
Oh those things also need to use cpu and ram.. charge.
Oh you want routes to destinations.. charge.
Literally anything you do in the cloud is monetized. If they did this with video games people would literally try to burn the buildings down.
That's doable, if your needs never change.I want the Earl Scheib one low price deal for my host.
yes but when you're getting the advertisement to choose a hosting service it is never spelled out that even moving data is charged for, having data is charged for, deleting data is charged for... as an example. You basically need to get procurement out to actually see if it's a good deal with IT people that ACTAUALLY know what's going to happen on the servers.Everything is monetized because everything has a cost. Hardware, software, licensing, rack space, bandwidth, power, man hours.
We charge for everything because we have to. Otherwise hosting is not profitable or sustainable.
yes but when you're getting the advertisement to choose a hosting service it is never spelled out that even moving data is charged for, having data is charged for, deleting data is charged for... as an example. You basically need to get procurement out to actually see if it's a good deal with IT people that ACTAUALLY know what's going to happen on the servers.
Because good lord doing big data or AI in the cloud is just bonkers expensive for anyone other than the cloud host.
Yea the cost up front is painful to build your own environment for the same task but over 3-5 years... yea... it's a different game altogether.
A great example is Lyft. They were 100% cloud until their cloud spend was such that to go on prem would be a cost savings.It's bonkers expensive for the cloud host as well. Which is why we charge a lot for the services. It's not a "once and done" purchase of the equipment. We get, maybe, 3-4 years out of millions of dollars worth of equipment before we have to replace it due to a 5 year EOL cycle. Then we have to buy it all over again, plus some. So we have to charge for the use of that equipment and services to recoup that cost, plus X% on top to account for projected increased equipment costs, plus staff, plus rack space, plus power, plus bandwidth, plus........
No one is going to invest and operate at a loss.
You'll pay a LOT for 99.999% uptime. That requires basically mirrored hardware, which very few do. Typical is maintaining N+1. And it still needs to be brought offline from time to time for maintenance/updates.A great example is Lyft. They were 100% cloud until their cloud spend was such that to go on prem would be a cost savings.
Also nobody in cloud offers five nines of guaranteed up time. My need for that is niche... Admittedly. But it's a need.
In a cloud sense, it really means a multi Availability Zone deployment, and ideally a warm standby in another region. I can count on one hand the number of times I’ve worked with a company willing to spend that much for 5 9s of availability.You'll pay a LOT for 99.999% uptime. That requires basically mirrored hardware, which very few do. Typical is maintaining N+1. And it still needs to be brought offline from time to time for maintenance/updates.
My companies environment I IT Engineer for specifically is decades of 100% uptime. Of course the redundancy level is beyond reason.You'll pay a LOT for 99.999% uptime. That requires basically mirrored hardware, which very few do. Typical is maintaining N+1. And it still needs to be brought offline from time to time for maintenance/updates.
I wonder how many cars have to go electric to compensate for the conflicts in Europe and middle east, I'm pretty sure those planes, tanks and rockets are also pretty bad for the environement, but who cares right?nd I get notices from my public utility telling me I use too much electricity running my box fan over night and that I should just turn the thermostat up even further...
There is literally zero way to be 100% uptime. Even with a full blown mirror image failover site you are beholden to the #1 killer of 100% uptime. DNS. And even then, in the event of a failover the transition is not instant. We had that setup. And it's expensive just to maintain. But, again, DNS doesn't propagate instantly. So even if we did fail over we're looking at a couple hours and up to 24 hours for propagation. Our clients would still be down regardless. We've had a couple outages, but none that lasted more than an hour or two over the past 12 years. So we ditched the fail over site.My companies environment I IT Engineer for specifically is decades of 100% uptime. Of course the redundancy level is beyond reason.
Each site is a active warm site for the other with replicated redundant systems at each site and redundant vmware clusters and the whole shebang. Basically everything in prod is at least 4x purchased simply to give us a reliable up time and purchased with integrated redundancies from network to power to power sources...
It's extreme but we wouldn't do it any other way.
I was talking about this earlier with some colleagues - we are all in the Power/Energy industry.
When it was Crypto - everyone was talking about how the massive power draw for mining was awful for the environment and how it's going to be the heat death of the Earth. I remember when the gas plant in upstate NY was purchased by a Crypto firm and everyone was talking about how it pollutes the lake.
But now it's not just some random Crypto Bros who are shucking out money to do this. Huge data centers, with Amazon/Microsoft money, are talking about nuke plants. And now everyone is just ok with it.
Even though all we did is change it from Crypto to AI - neither of them really serve useful purposes right now. And both take massive amounts of energy.
And I get notices from my public utility telling me I use too much electricity running my box fan over night and that I should just turn the thermostat up even further...
Man we need to get the big power facilities online. Nuclear is the way to go for that fully. Without a doubt. Power at scale easily sustainable. no longer really needing to deplete uranium or the like.SMR's are coming. There's no other option for large scale, rapid deployment of energy infrastructure. Plus it distributes the load across the grid instead of being centralized as with a large plant.
That's what my wife does, btw. Nuclear development. It's a shame the U.S. is 20+ years behind every other developed country. Clean, consistent power, but the NIMBY's lose their minds over it.