I don't understand how it takes 3 years to get off the cloud. I'm not a cloud developer, though. The most I've done is run code on free hosts or compute instances. Presumably there's something to the microservices and lambdas and distributed compute that makes this hard. I'm thinking if this was a monolith (like AWS themselves admit is cheaper), they could just run it locally? What a giant waste of money. I'm very glad to start seeing xAAS start to die out. At the end of the day it's just looking like more middle-men instead of how I've always assumed it was intended to be: economies of scale.
However, and missing from this article + discussion so far, is their revenue. If they pay $4/day and make $2 in revenue, that's bad. They pay $300k/day but make ~ $2250k/day in revenue. I don't know what the ratio is supposed to be, but at first blush that doesn't actually seem too bad. I'll let the more qualified take over, I'm struggling to find out how big a % of their total expenses this is.
If you use cloud brand queue system, you not only need to run some servers to run the queue, you also need to rewrite all parts of the system that relied on cloud brand queue system into something else.
Repeat it by a bunch of cloud services your app use and you're in for major rewrite.
Cloud offers many, many "good enough" solutions for common problems in development and it is oh so easy to use another cloud api for few cents more
And that's not even touching a bunch of auxiliary services you now need to provide like monitoring, backups, metrics etc.
Usually best strategy is to start move from most expensive parts but if that part is entangled in cloud services it could take forever
It's insane spends like that which are the main reason that reducing cloud costs is the main money maker for my consulting business at the moment.
It's extremely rare to come across people using cloud services where you can't cut their costs by at least 30%, often more than half, sometimes as much as 90%+, and usually it results in reduced devops spends as well.
The interesting thing is that the hardest part of the sales process is to convince people that cloud isn't cheap, because there's a near religious belief that cloud providers must be cost effective in some companies.
I've seen so many people being baited by whole "you will not only not need to buy expensive servers, but also not have expensive ops guys to manage it!" schtick.
Meanwhile your devs put more man-hours than your actual hardware ops teams would ever do into wrangling cloud quirks and spend way more than getting a bunch of servers (even if you include costs of financing them in that) would ever cost.
It’s wild how someone can invent an idea, force it on everyone something no one asked for then turn around and sell it to us, layering on FOMO to keep us hooked.
Emotional manipulation at its finest.
No wonder nothing feels special anymore. Everything’s automated, abstracted, owned by someone else. If you make things by hand, you have to prove their value meanwhile, platforms like AWS are printing money.
They want us to own nothing. And if “heaven” is real, why aren’t they there already?
This is basically technical debt of some sort - they may be minting so much money it doesn't matter atm, but eventually, someone will get some wins cutting costs. Probably just the bandwidth is a lot of it.
I see lot of comments questioning why can't they move to another provider. When you start using any of the cloud providers there so many services you start using and then you start building integrations, email notifications, api gateways, and what not. This makes it very difficult to move all the established processes.
This is why I don't use AWS or any "cloud" provider. That is not 100% true. I have 110 GB or so of personal data backed up on Glacier. The bill is 6 cents a month, and they obviously don't charge my card.
It may take longer to start, but owning all your servers is cheaper in the long run.
At the very least, don't lock yourself into one provider. Spread out any services you use.
More than one company has screwed itself over permanently by tying itself to a single service. At the end of the day, it is the extremely poor decision to follow the trend of locking oneself into a single company that dooms it.
You would think people would have learned the lesson to avoid lock-in by now. It is absolutely Darwinian at this point.
17 comments
[ 3.5 ms ] story [ 82.1 ms ] threadHowever, and missing from this article + discussion so far, is their revenue. If they pay $4/day and make $2 in revenue, that's bad. They pay $300k/day but make ~ $2250k/day in revenue. I don't know what the ratio is supposed to be, but at first blush that doesn't actually seem too bad. I'll let the more qualified take over, I'm struggling to find out how big a % of their total expenses this is.
Repeat it by a bunch of cloud services your app use and you're in for major rewrite.
Cloud offers many, many "good enough" solutions for common problems in development and it is oh so easy to use another cloud api for few cents more
And that's not even touching a bunch of auxiliary services you now need to provide like monitoring, backups, metrics etc.
Usually best strategy is to start move from most expensive parts but if that part is entangled in cloud services it could take forever
Seriously the CTO should be fired.
It's extremely rare to come across people using cloud services where you can't cut their costs by at least 30%, often more than half, sometimes as much as 90%+, and usually it results in reduced devops spends as well.
The interesting thing is that the hardest part of the sales process is to convince people that cloud isn't cheap, because there's a near religious belief that cloud providers must be cost effective in some companies.
Meanwhile your devs put more man-hours than your actual hardware ops teams would ever do into wrangling cloud quirks and spend way more than getting a bunch of servers (even if you include costs of financing them in that) would ever cost.
Some web sockets with CRDT processes running.
What else?
Emotional manipulation at its finest.
No wonder nothing feels special anymore. Everything’s automated, abstracted, owned by someone else. If you make things by hand, you have to prove their value meanwhile, platforms like AWS are printing money.
They want us to own nothing. And if “heaven” is real, why aren’t they there already?
It may take longer to start, but owning all your servers is cheaper in the long run.
At the very least, don't lock yourself into one provider. Spread out any services you use.
More than one company has screwed itself over permanently by tying itself to a single service. At the end of the day, it is the extremely poor decision to follow the trend of locking oneself into a single company that dooms it.
You would think people would have learned the lesson to avoid lock-in by now. It is absolutely Darwinian at this point.