Cloud is here to stay, but customers are starting to question the cost Cloud-based infrastructure services date back at least as far as 2006, when AWS introduced its S3 storage platform, followed by Elastic Compute Cloud (EC2) instances. Since then, cloud has become a global industry topping $100 billion in size, but some customers have begun to question the move to these services and started to …
when "offshoring " was all the rage. Right up until (UK) companies pissed away their in house resources and suddenly had to pay YoY increases of 10-15%. Which not only wiped out any "savings" made by the sackings, but made IT twice as expensive as before.
Now it's "cloud". Yeah, Lets hive off business critical functions to a bunch of folk who (if they stay solvent and interested*) will eventually be able to jack the subscription up every month if they like.
Recently I have pulled a series of eye-watering cloudy services back into on-prem/hosted servers where we run them. Currently savings are £11.000 a year.
Our CEO occasionally mingles at events. Some outfits are quietly regretting losing that resource.
I'm thinking the next few years will be good for people who can spin up a LAMP/Docker server, and plumb it into a company to replace ever-rising cloud costs. A job which can be done 100% remotely.
I hope so, it may just be my retirement plan.
*Not quite in this vein, but Stackpath deciding this ain't for them is another risk you have with cloud. Or things like Amazon retiring MWS with little notice.
In local government, it's usually about having someone to blame if it goes wrong - preferably a supplier who you can drag through the newspapers and/or courts.
"We're a council and committed to saving taxpayers' money, while they, your honour... (grinds teeth in anger)"
Yeah, Lets hive off business critical functions to a bunch of folk ...
When the only driving force behind this type of business decisions (if we can call them that) is just making more money, this is the result.
Obviously applies to both sides: the DHs that offshored and lost their most valuable assets and the DHs that thought they could fleece their clients without consequences.
.
As usual, blind zeal in technology leads once more to "all or nothing" strategies. Cloud is good for some things and not so good for others and the pitiful excuse you can't control costs is BS, you can control the costs of anything you simply have to put the leg work in and automated reports. For example where I am if anything spins up the monthly cost is immediately calculated and if the costs is about a set threshold it fires out alerts warning of over-spend. We've linked the IAM controls on our clouds into into AD system so only about 10 people out of 200 have any access to cloud tech, if you need access then you submit a proposal to a committee and they decide if you have grounds for access.
Heck we even have two guys who spend 1/3 time tracking cost and usage in our clouds, we nicknamed them "The KGB" as they appear without warning when you so much as click on a cloud web interface! Ha ha!
I still think there's a lot of mileage in on-prem DCs, however people are trying to apply the same opexe/capex models to everything, on-prem certainly doesn't work like that you have spend up front, get vlue and then save and prepare for the next update in 2-3 years time. In cloud you allocate a max budget, broken down by resource type and you monitor and control.
There's a lot to be said for keeping your baseline infrastructure in-house to contain costs, but using the hyperscalers for... when you need to scale up in a hurry. It's also easier to do new development and run experiments on a hyperscaler because you can provision what you need and then switch it off when you're finished. But it does, as you say, require discipline. And education. Left to their own devices users will provision large and expensive compute instances they don't need, and then leave them running for months.
It does not need to be all in or all out - it can be hybrid, too :-)
The other thing is that considerations as to where you put your compute power depends on many factors - cost is only one of them.
Probably best for businesses not just to follow a general trend of cloud vs. on-prem, but do some thinking and analysis to see what is best - now, and for the future.
"Over-provisioning of resources and idle or underutilized resources affected half of all respondents, with lack of skills or not having the right capabilities to manage resources blamed."
I'm reminded of the mustard manufacturer who said his profits come from the mustard diners left on the side of their plates.
Too late. They already have you by the balls, don't give two hoots about your hearts and minds, and are at liberty to tighten the pincers whenever they choose. When you place your mission critical resources in the care of those who are only interested in your wallet you get what you should have expected. What a pity forethought is not more prevalent.
...many have questioned the cloud providers' nebulous mathematics about how exactly it's cheaper to do business with them. At the end of the day it all comes down how much room do the customers have in their budgets. At this point cloud providers are raking in so much cash that all they have to do to endure a once in a century recession is stop innovating for a short while. They wouldn't even have to lay off anyone unless they wanted to but instead just shuffle their talent to to fit the environment. This should be a signal that most things in the cloud are still over priced. However cloud customers keep letting themselves be charged more because although cloud is expensive it's easier than back tracking on your plans to use cloud resources. If cloud customers slashed their shopping budgets by 50% it's wouldn't be the end of the cloud, but it might be the reckoning the cloud deserves.
"Part of the issue seems to be that cloud is not easier to manage than looking after your own infrastructure, but instead presents a different set of management challenges."
For a small to medium sized organisation which has requirements beyond SaaS, it's worse - Azure is designed for large enterprises and unlike on-prem server infrastructure it doesn't scale down - all the functionality and configuration options remain there, making it massively complicated for an organisation just requiring a basic network of say a dozen or two VMs. It's also a minefield as regards price traps, with default options selected which may well be more advanced and expensive than required, but unless whoever is setting it up knows all the functions well they aren't going to necessarily know this. I've not really used the other major IaaS services but I assume similar will apply?
I can only speak for a large org with respect to hidden costs. Here's a couple of examples from the coal-face:
1) A number of business cases were made on the basis of securing "Spot Instances", which failed in practice because there wasn't sufficient liquidity to meet demand. So the applications were *forced* to use "Reserved Instances" instead - which were found to be ~3x more expensive than on-prem hosting.
2) Application teams had to acquire and/or develop system administration capabilities, duties that were previously handled by on-prem hosting teams (at lower cost).
Also with respect to "unused" capacity cloud doesn't really tackle the root cause of that - specifically the fear of "if I switch this off what else is it going to take down ?". It's not necessarily a rational fear, but it exists with cloud infrastructure just the same as it does with on-prem.
I have been using and writing about a wide range of cloud technologies from their beginning. I have coined a term to describe what cloud usage has become and that is "Labyrinthian ". All public clouds have become a labyrinth of interdependent settings where the above comment about switching something off may inadvertently take something else down as well is germane. The new promise is that AI will help with all this and assist our puny mammalian brains with all the complexities. However, I am quite confident that I have never hallucinated any of my cloud settings, or deletions, but I am quite confident that AI can and will do so. That observation comes from using LLM's daily since they first arrived on the scene. You can be absolutely assured they will seriously mess up some cloud configuration with unusual complexity and multiple steps that have to happen over time. It will be akin to telling your senile grandfather where and when to click for the entire process to be successful. Further, I have been consuming all the information I can gather about Microsoft CoPilot for the enterprise. With regards to being Labyrinthian, you ain't seen nothin yet. There will be articles here where people try to best each other on the colossal nature of the mistakes it made. Things like " all we did was delete an old Sharepoint folder and it sent emails to all our customers about how we were no longer accepting orders." Fun times will be had by all after you've spent over a year implementing it. Oh, and don't worry if you are under 300 seats. it's not for you. That could actually turn out to be the grace of capitalism..
As ever, its "right tool for the job". Sometimes it is on-prem and sometimes it is the cloud. But just because it is in the clould you still need (among a long list, lets not nitpick here) SME, failover, DR, redundancy, backups, security, monitoring, compliance, testing, debugging, capacity planning, cost monitoring, etc, etc. NONE of that is magic'ed a away just because you're in a cloud. And its all expensive.
On prem is a skill set. Read, train, practice. Break it and fix it.
Cloud is a skill set. Read, train, practice. Break it and fix it. Just like you learned with your crappy 486 back in the day.
There are some upside though. Before my company went all-in on the cloud, there were switches in offices that were old enough to drink in the US. Always put back to the next budget cycle. That is a much managements fault for not understanding that this stuff isn't going to last for ever, as it was IT fault for not selling the risk and consequences. Now, the cost is rolled up. Still, management hasn't got why on-prem performance is (still) much worse that via VPN.
"operating applications and services in the cloud can be just as costly as owning and managing your own infrastructure for this purpose, and sometimes more."
#insert <homer simpson voice.h>
There's little savings when all you've done is move apps and services from self-managed (on-prem) to the Cloud! Someone still has to build, deploy and keep them running which I said back in the early naughties!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
