Why Companies Are Leaving the Cloud (infoworld.com) 176
InfoWorld reports:
Don't look now, but 25% of organizations surveyed in the United Kingdom have already moved half or more of their cloud-based workloads back to on-premises infrastructures. This is according to a recent study by Citrix, a Cloud Software Group business unit. The survey questioned 350 IT leaders on their current approaches to cloud computing. The survey also showed that 93% of respondents had been involved with a cloud repatriation project in the past three years. That is a lot of repatriation. Why?
Security issues and high project expectations were reported as the top motivators (33%) for relocating some cloud-based workloads back to on-premises infrastructures such as enterprise data centers, colocation providers, and managed service providers (MSPs). Another significant driver was the failure to meet internal expectations, at 24%... Those surveyed also cited unexpected costs, performance issues, compatibility problems, and service downtime. The most common motivator for repatriation I've been seeing is cost. In the survey, more than 43% of IT leaders found that moving applications and data from on-premises to the cloud was more expensive than expected.
Although not a part of the survey, the cost of operating applications and storing data on the cloud has also been significantly more expensive than most enterprises expected. The cost-benefit analysis of cloud versus on-premises infrastructure varies greatly depending on the organization... The cloud is a good fit for modern applications that leverage a group of services, such as serverless, containers, or clustering. However, that doesn't describe most enterprise applications.
The article cautions, "Don't feel sorry for the public cloud providers."
"Any losses from repatriation will be quickly replaced by the vast amounts of infrastructure needed to build and run AI-based systems... As I've said a few times here, cloud conferences have become genAI conferences, which will continue for several years."
Security issues and high project expectations were reported as the top motivators (33%) for relocating some cloud-based workloads back to on-premises infrastructures such as enterprise data centers, colocation providers, and managed service providers (MSPs). Another significant driver was the failure to meet internal expectations, at 24%... Those surveyed also cited unexpected costs, performance issues, compatibility problems, and service downtime. The most common motivator for repatriation I've been seeing is cost. In the survey, more than 43% of IT leaders found that moving applications and data from on-premises to the cloud was more expensive than expected.
Although not a part of the survey, the cost of operating applications and storing data on the cloud has also been significantly more expensive than most enterprises expected. The cost-benefit analysis of cloud versus on-premises infrastructure varies greatly depending on the organization... The cloud is a good fit for modern applications that leverage a group of services, such as serverless, containers, or clustering. However, that doesn't describe most enterprise applications.
The article cautions, "Don't feel sorry for the public cloud providers."
"Any losses from repatriation will be quickly replaced by the vast amounts of infrastructure needed to build and run AI-based systems... As I've said a few times here, cloud conferences have become genAI conferences, which will continue for several years."
Just wait, de-clouding isn't the end of it. (Score:5, Insightful)
One day, managers will realize that if you have to hire IT people, there's no difference between Linux and Windows in terms of 'knowing it' because you hire people who already do.
One day, managers will realize that a Linux desktop is good enough, and no, you don't really need all those extra things shoved into Windows that Microsoft changes every few years anyway. 99% of what goes on at the desktop level these days is checking email, instant messaging, and maybe very basic MS Office suite. A lot of business apps are now just front ends connecting to a web server.
Companies already have 'those Macs that are special cases', I see no reason why you wouldn't have 'those Windows systems that are special cases' and make all the common use stations Linux. Imagine how much less money you'd have to send to Microsoft for OS and 365 subscriptions.
But more on topic... as for 'the cloud'? Bah. Never liked it. Host your own mail. Host your own web server (for most companies, their web site is a glorified vanity page). Contract with a local IT company to keep it running if you're too small to have your own staff. The only thing that should be hosted is your off site backups.
Re:Just wait, de-clouding isn't the end of it. (Score:5, Informative)
Re:Just wait, de-clouding isn't the end of it. (Score:5, Interesting)
Why windows over Linux in many business environments?
2 reasons:
1) most staff already know windows. Yes, this is a chicken and egg problem for corporate desktop Linux. It is what it is.
2) Microsoft Office. Specifically Word, Excel, PowerPoint. People know how to use them and expect them to work as they do. Anything that varies won't be adopted. No one wants to learn a new app to get the functionality they already had.
1 not reason:
1) costs. The cost of the OS is already wrapped into the hardware. No one cares. There's a budget. It must be spent or risk losing the unspent in the next cycle. My personal target spend was 1% more than my budget. The CFO praises me for hitting target, doesn't freak about about 1% over when others are 50% to 300% over, my budget is never cut and I can always ask for extra if I have a large project out of normal budget scope because I got target consistently.
When wearing my IT Manager hat, Linux did absolutely nothing to improve my life.
Linux was in the data center and I did avoid several jobs that had critical production windows data center infrastructure but for the office, windows all the way.
Re:Just wait, de-clouding isn't the end of it. (Score:4, Interesting)
Every major Windows version change, I hear people complain about the desktop differences... and it seems to last until the next major change. Remember when the Ribbon became a thing?
I run default Mint on a couple of computers in my home and my wife, who is almost completely computer illiterate, has no issues switching between it and Windows. We use Bluemail as an email client and sometimes the interface annoys her, but then so do Hotmail and Outlook. I know several companies that somehow manage to use a Teams alternative - Windows based, but not Teams. Every VOIP softphone is a little different, and people manage.
I no longer see an insurmountable issue retraining an average desktop user doing average things from Windows to some carefully chosen variant of Linux. It's no longer significantly more to ask than adjusting to a new Windows interface.
Re: (Score:2)
This is why enterprise applications are on stable LTS. Eventually, you have to update and retrain, but screwy SaaS variability is for the consumer. In an enterprise or small business setting, you gpedit all that crap into oblivion or you don't even get it. Businesses were on Win7 for years. Now they're on Win10 for years. Next time a switch will happen will probably be Windows 13 (which will be called something else because... 13. Maybe FFS?). Maybe the version after that (14?), if 12 is really coming out t
Re: (Score:2)
Every major Windows version change, I hear people complain about the desktop differences... and it seems to last until the next major change.
Exactly. Even a windows update causes issues for end users. What do you think an OS change will do?
I've been using Linux full time since 2001, I'd run my own servers, served as the backup admin for more than one employer, I'm not a guru but I'm easily in the 10% of whatever group of software devs I'm in.
And I spent two hours this morning unsuccessfully trying to get my ubuntu system to automatically choose the right output device after I wake/boot my computer.
Remember when the Ribbon became a thing?
I run default Mint on a couple of computers in my home and my wife, who is almost completely computer illiterate, has no issues switching between it and Windows.
Your wife has a dedicated IT person if she has a
Re: (Score:2)
I mostly agree with you here but this doesn’t really address the point that microsoft will happily fuck with everything for silly reasons. Unless you use extremely niche features libreoffice is closer to office 2003 than the ribbon offices
Re: (Score:2)
The big problem with LO for me is Calc. I use Excel at work fairly regularly and it is just less painful to use in honestly every single way. The interface is more comfortable with fewer clicks and keypresses required, the onboard help is excellent, the context help likewise, and perhaps more important than any of those things it's very noticeably faster in every regard. The functionality is also light years ahead, e.g. live pivot tables. I'm thankful to have a free of cost office suite that does a reasonab
Re: (Score:2)
LO Calc has some ridiculous bugs and missing features. I gave up and went back to Excel. Additionally, when I report a bug they tell me it isn't a bug, so clearly they will never fix some of these things.
Re: (Score:2)
Right. But guess what is even closer to Office 2003 than LibreOffice? Yup, you guessed it. The answer is Office 2003. Which is why I keep using it. There are very few features I ever need from later versions. True, Office 2003 has a few annoying bugs. But LibreOffice has an enormous quantity of annoying bugs that have been deal killers for some things I've tried to do, like stuff as simple as putting images into a document.
Re: (Score:2)
I have to admit I keep a copy of office 2003 as well but keep it confined to a few use cases because it has so many unmitigated CVEs. It’s peak office and even if you go back a few editions it only has a few compelling features (aside from docx support)
Soooooo fast on modern machines.
Re: (Score:2)
For someone who is 'way smarter than me', you seem to have a lot of trouble grasping the underlying point that there isn't much difference between retraining between Windows versions and retraining to one of the friendlier Linux distributions.
You keep saying, "It's different". It's not. There's always retraining and documentation regardless of whether you stay with Microsoft or not. Are all your users still trained on and using documentation for Windows 3.11 for Workgroups?
Re:Just wait, de-clouding isn't the end of it. (Score:4, Insightful)
There is a huuuuuge difference between training someone to use Linux and the next version of Windows/Office.
The significant changes to windows are very rare and not a stumbling block to half the users.
Switching them to Linux is definitely harder, requires a complete rewrite of all documentation and -still- has absolutely no benefit to the company.
What is this amazing as yet to be named benefit to switch to corporate desktop Linux/office?
Literally no one in corporate gives a shit about the extra $100 pass through for MS Windows or Office license. Larger places just buy an enterprise license anyway. Still waiting to hear from -anyone- why desktop Linux is a better option.
Show me how you're smarter than me. What is this mysterious benefit?
Re: (Score:2)
Using the DMV example, it makes me wonder why they got rid of the 3270 terminals. For a lot of things, especially forms that needed to be filled out, nothing beat those for being effective, as one winds up having to tab through all the fields anyway. Maybe some better handling of complex, hierarchical menus like finding a product, but overall, from what I've seen, moving to Web based terminals has made things closer than just going through fields.
OS-wise, going from skeuomorphic UIs to flat-file with butt
Re: (Score:2)
2) Microsoft Office. Specifically Word, Excel, PowerPoint.
95% of people can use Google Docs for all of those.
It's free, accessible anywhere, works in any browser, most students use it in school, and Google Docs makes collaboration easy.
Re: (Score:2)
Re:Just wait, de-clouding isn't the end of it. (Score:4, Insightful)
Re: (Score:2)
Re: (Score:2)
I go for Linux for most of my work (mainly thanks to Plasma giving me window management features I can't get in Windows), but still need a Windows VM upon occasion:
-I use Teams via browser 99% of the time, but if I need to control someone else's screen share, Microsoft requires the desktop app, so to Windows.
-On the occasion I need to touch a powerpoint made using company template and have to change a part, I need the Windows VM because web powerpoint won't let me touch any master slide content
-Once in a wh
Re: (Score:2)
You missed on reason, MS integrates a lot of their alleged software. You might not like how it integrates but there it is. DoD is going through this now and forcing MS crapware everywhere whether it makes sense or not.
There is an unspoken reason too: When a ship was immobilized awhile back from using Windows NT and had to be towed to port, one Naval guy was asked why use something as dicey as Windows (not sure LInux was really a viable option back them, but Unix was). The Naval guy said when they had really
Re: (Score:2)
The time period you’re referring to, holy crap the military was AWFUL with computers. So much day to day shit was run by teams of enlisted people you’d have leadership that spent their whole careers checking diodes leading kids who had a few months of school. By the time I got out they were all talking about low level comptia certs like they were gold.
It definitely seems to have improved but microsoft must have booth babes giving out BJs to senior officers or some shit because the DoDs appetit
Re: Just wait, de-clouding isn't the end of it. (Score:2)
Re: (Score:2)
As a consequence the company moved everything to Windows.
Since I am retired I'm working as contractor for a couple of companies and run all the office stuff on Linux.
LibreOffice just works for text, not too complicated spreadsheets and presentations, Teams, Outlook and the agenda work fine in the Chromium browser. Sure, MS stopped supporting the Teams version for Linux but the web version works fine.
Yes the companies pay
Re: (Score:3, Interesting)
I'm not a Linux zealot, I'm a miser. I am a Windows IT tech by trade and a Linux user at home. Almost everything that work pays for, I do for free at home. The things I don't do for free at home generally are things the average user complains about and rarely uses. The things that I can't do for free at home are usually web apps these days.
You can migrate away from Windows. And if you want vendor support, there are in fact vendors who offer paid support options for Linux distributions and there are app
Re: (Score:2)
I have a couple usability issues* with Mint at home that would likely keep it from being recommended at work, but I agree that the subscription model does make alternatives more appealing. I switched to LibreOffice and it works fine for me, although the font situation is complicated.
* Keyboard focus problems mean that the right spot in some programs do not take text input without moving the mouse. This is painful for things like CAD where you type a lot of commands for speed.
Re: (Score:2)
I've never seen companies train users how to use their desktops, they just get what they're given and are expected to get on with it on their own.
Re: (Score:2)
Re: (Score:2)
What training? Using a word processor or a spreadsheet is as different between MS version updates as it it between vendors. The reality is most office workers know the basics and never use the advanced features that may be different. People just keep spewing this drivel.
Then you have MS pushing their junkware and breaking workflows like the HP driver update or one drive. Microsoft zealots can't even admit the problems and just invent excuses. Printing was solved on every other platform 20 years ago when usi
Re:Just wait, de-clouding isn't the end of it. (Score:5, Interesting)
99% of what goes on at the desktop level these days is checking email, instant messaging, and maybe very basic MS Office suite.
I don't think it's that as much as the fact of all that can get wrapped up in Active Directory and Group Management which if you're an IT department tasked with managing hundreds or thousands or tens of thousands of user machines is something they are probably wary of giving up, especially on the Enterprise level where paying for Microsoft is kinda baked into the business
Re:Just wait, de-clouding isn't the end of it. (Score:4, Insightful)
Host your own mail. Host your own web server (for most companies, their web site is a glorified vanity page).
Have a process where you can patch them quickly. That is the most important part.
Re: (Score:2)
Re:Just wait, de-clouding isn't the end of it. (Score:4, Informative)
Re: (Score:2)
I’m not sure what they see in it but nontechnical management will never give up their shitty microsoft apps. Multiple times I’ve had to defend spending money on absolutely critical shit that would have sunk the org in a year or less if we cancelled. But in my life I’ve never ever seen a manager question the value of a microsoft product.
Re: (Score:3)
Companies already have 'those Macs that are special cases', I see no reason why you wouldn't have 'those Windows systems that are special cases' and make all the common use stations Linux. Imagine how much less money you'd have to send to Microsoft for OS and 365 subscriptions.
There's the rub, though. Businesses need Office because the other businesses use Office, and subtle problems occur when loading the baroque Microsoft formats into other applications. So they're not likely to get away from the 365 subscription, though yes they could and should get away from the OS as Windows 11 is miserable.
A lot of businesses also seem to stick with Windows because of the one system for centralized management. This seems like the most compelling argument for keeping it, you can find someone
Re: (Score:2)
they could and should get away from the OS as Windows 11 is miserable.
Speaking as a Linux user at a MSOffice-centric company, not really. If suddenly they want me to control someone's screen in Teams, they have to wait while I boot my Windows VM, because Teams for web doesn't have that. If I have to deal with a powerpoint, there's a high chance I need to fire up the Widnows VM, because the web version is missing a lot of capability. Despite their impressive web support, ultimately they still have Windows applications as the fully enabled facility.
Re: (Score:2)
Edge is just bad Chrome, Chrome has remote desktop, realistically you'd standardize on Chrome on your Linux clients and you'd have that functionality.
Yeah, the stuff missing in the web versions of office apps is an issue now, but for some reason they are going that way and those issues will be addressed. A lot of places are going web-only so they don't have to install apps on PCs.
Re: (Score:3)
Except I'm not using Edge when I'm in Windows, it's the native app. And sure, they could install the remote desktopp chrome extension, or teamviewer, or anydesk, or any number of similar stuff, but particularly when dealing with other companies, they want me to control on their terms. I don't get to dictate their solution. So if they use Zoom, great I'll use Zoom, and use it from Linux. But if they use teams (and 90% of them do, because it comes "free" with their O365 subscription), then I have to fire up
Re:Just wait, de-clouding isn't the end of it. (Score:4, Interesting)
Unfortunately, no, it wouldn't play out that way.
If they are so "don't care about Windows", then Windows still comes "free" with the device purchase. So the path of least resistance is to just go with Windows. It would be more likely that a company decides to eschew the concept of company owned personal computers and just have the employee take care of it, if they decide it's all just a web browser anyway.
If they care about Windows, specifically, then they are likely attached to a concept that's pretty unique to Windows: the physical owner of the device can be blocked from making their own decisions and be audited more easily. Things like locking down browser settings or even changing the wallpaper on their users desktop. Also, all the various anti-(malware/ransomware) solutions they have been convinced are absolutely needed are unlikely to support Linux. Trying to make the argument that Linux shouldn't need it will not get too far, because the same rationale for that should apply to modern Windows, in theory. Besides, there have been Linux malware that has run over lazily administered Linux instances.
O365 remains a powerful motivator with strong networking effects. If you interact with another company and have a risk of a pptx misrendering becuase of your use of libreoffice, well, that can look "unprofessional" and undermine confidence in your company. A mitigating factor is that the subscriptions come with web access that is generally serviceable, so maybe you can get by with Linux. Except when you can't, because at least PowerPoint and Excel are missing a lot of features. If II have to update part of the company template, I likely can't because I need master slide view which isn't in the web version.
I'll confess to be along your mindset for cloud hosting, but I can recognize that some companies will have online services that they want to provide that would be more robust than their own internet connection, and more local than reaching out to their own infrastructure. On the other hand, on premise hosting means the services *locally* are more robust than your internet connection, and faster with more predictable costs.
Re: (Score:2)
Yep, pretty much. But managers come with huge egos and tiny skills, so this may still take a while. I recently saw that they now teach "evidence based management" so the field may slowly become more professional.
Re: (Score:2)
Host your own mail. Host your own web server (for most companies, their web site is a glorified vanity page). Contract with a local IT company to keep it running if you're too small to have your own staff. The only thing that should be hosted is your off site backups.
Re: (Score:2)
If KDE and others GUI interfaces want to get used by normal people. They need to be designed in that way. Currently that's not the case. For instance, using a WiFi network in Linux with KDE requires (this can maybe be turned off) KDE wallet to be used before the WiFi network connects. This is an extra step and it is not needed. It just bothers users, is off putting and complicated the whole user experience. KDE and other GUI in Linux are full of this un-needed things, extra steps that are not needed.
In Wind
Re: (Score:2)
Host your own mail. Host your own web server (for most companies, their web site is a glorified vanity page). Contract with a local IT company to keep it running if you're too small to have your own staff. The only thing that should be hosted is your off site backups.
Let's try that again...
If you're small, you should absolutely not host your own email and web server. You will have to deal with a deluge of spam. You'll have to keep up with security of your server on the internet, including monitoring. To do that, you need an experienced, dedicated staff. If uptime is important, you'll have to spend on providing redundancy.
Better to outsource email / web to a provider that has teams dedicated to each item that will likely be better at it than your IT team. Sign a SLA
schadenfreude (Score:5, Interesting)
Re: (Score:2)
As someone who made a little cash selling "consulting" to wannabe "movers to the cloud", I'm amused too.
Now I could charge for the reverse move if I were still in that business :)
Re: (Score:2)
If I were the person who made decision to move, sure. But I'm not, so the most they'll get if they sue (not that they'd want to) is to pay my defense costs.
Re: (Score:2)
And yet it was advice that, at the time, was considered to be correct.
Re: (Score:3)
A consultant would implement a cloud solution, not recommend it. We're very careful about letting people know what their options are and letting them pick their poison.
Obvious issues seem obvious (Score:5, Interesting)
A lot of people in positions of power don't seem to realize that a move to the cloud means you still need people to manage those cloud servers / services - or else you have to pay the cloud companies (or contractors) a significant sum to do it for you. Or, for that matter, that you might need to also think about stuff like bandwidth and just how much CPU time you'll need.
I work in a university STEM department. You'd think STEM faculty would understand this, if anyone does, but - for a while the department's leaders were pushing to move everything "to the cloud" because they thought it would save them tons of money. Mainly because all they looked at was the monthly charge for no-support server instances and thought "Hey that's gonna save us a bunch since we won't have to pay IT salaries anymore".
Eventually they got enlightened by an outside contractor they hired to document just how much money the move would save (after a handful of faculty said - of the original plan - "I don't think this works the way you think it works").
Re: Obvious issues seem obvious (Score:2)
Re:Obvious issues seem obvious (Score:5, Interesting)
I've had to relay the same "revelation" to those in charge. You often don't save money. You actually spend more, but you get more capacity for scalability and redundancy... if you're willing to pay for it. Most workloads for internal company use don't really need that.
Re: (Score:2)
In addition, for my business, we also get to offload liability to the cloud provider.
Re: (Score:2)
Well we're already doing that with a managed hosting provider. If something goes wrong, just blame them. Works like a charm.
Re: (Score:2)
I agree,
Cuts to IT when cloud resources need management and maintenance cause some big blowouts in cloud subscription costs.
I still see to this day the non-sensical lift and shift being a goal of Enterprise organisations. Expecting that cloud will solve ever growing costs of maintaining everything. As application usage, data rate, and data storage increase if your crude infra that you uplifted to the cloud doesn't adapt it's cost are going to spiral.
You can't treat cloud as simply on-prem virtualisation.
S
Bad cloud use cases killed cloud (Score:5, Interesting)
Cloud is great for bursty loads and short term projects. In my experience most loads don't fall into these categories.
Cloud is terrible for near constant loads over long periods of time. Data center is way cheaper and easier to run.
Cloud management requires a level of OCD control that most places can't handle. It's too easy to spin up expensive virtuals that get lost, burn through expensive storage you didn't realize you were using, not realize your outbound data has spiked into the danger zone until the bill comes due, etc. And there goes your budget in a giant flaming wreck.
Initial setup with your tool of choice is also non trivial and requires time, experience and effort to get right.
Data center is impossible to accidentally over use your resources. You own what you own. Set up monitors for network, disk, cpu, etc, and get alerts when things start to go bad. Straight forward and no budget risk from run away jobs. Harder to lose track of some random server when you have a spreadsheet of everything you bought and what it was assigned to. You can do this in cloud but the creation/destruction of vms is often automated so you don't really know what you have at any moment. Too dynamic in many environments. Boom, there goes your budget again if you blink.
The down sides to data center is initial setup. You have to know your hardware requirements, networking, storage, etc, before you start. But once you have a footprint in place adding more servers is straight forward. Your long term data center cost for the flat load is way lower than cloud. Winning for that common use case.
I was on way too many projects where the new CTO hired me to come in to move his predecessor's data center to cloud even though it made no technical or financial sense. Totally stupid. But I sure made bank doing it. I can't recall a single data center -> cloud migration project that should have been done.
Re: (Score:3)
I am curious to hear more about your experiences. Was it your sense that the "new CTO's" (A) had done any due diligence deciding to move to the cloud; maybe coming up with the wrong numbers or answers, but at least an honest effort; or (B) they were prodded, goaded, badgered into doing so by the CEO or CFO or someone else higher but more stupid, so they dutifully complied; or (C) they were just jumping on the latest tech fad bandwagon, IT-fomo; or (D) something else? Did you ever get the sense that the
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Well tired, too, yes but fired. :-)
I always understood my role was to get paid to take the heat for his failures after fixing them for him.
Re: (Score:2)
Harder to lose track of some random server when you have a spreadsheet of everything you bought and what it was assigned to.
I lost a RPi in my house for a couple weeks. It worked, was connected, but I had forgotten where it was physically. Turns out it was in my attic, and when I had moved a bunch of hardware and networking equipment from there, I had overlooked it because during the moving activity, it fell behind the small rack I had there (which I left in the attic "just in case").
Yes, I know there are ways to find it, and that's how I figured it was there, but that was only because I wanted to increase its storage, otherwise
Re: (Score:2)
Re: (Score:2)
An old company department I worked at had a room for datacenter use as we were super adamant about on-prem.
Hardware purchases were next to impossible. First anything physical and company had to approve and then inventory which was hard. But, that was the smaller problem. The bigger problem was someone to setup the hardware and not have it interfere with already what was already running. IT was super-reticent about changing anything and could always bring up thousands of excuses not to do any changes.
Our tim
Re: (Score:2)
There are always shitty groups in every company and unfortunately an out of date stodgy IT group is one of the worst to have.
Cloud has a role and purpose. Non-cloud of all types has a more common role and purpose. Given my choice and reasonable budget I usually blend both in a hybrid model to get the best of both world and put the right jobs in the right locations.
As you noted your IT just sucked. Sorry about that. They make the rest of us look like shit. Everyone isn't like that.
Apparently everyone th
Re: (Score:2)
The cloud is far from dead. The story notes that 25% of companies in the survey were moving back to on-premise. That means that 75% were either moving TO the cloud, or NOT moving off of it.
two words (Score:2)
egress costs.
Cloud and remote work (Score:3)
Cloud services are natively "remote". It seems a good fit for remote workers.
Re: (Score:2)
That's not what the article is about. You're looking at it from the user's point of view: to you, using service X from your home is easy because it's SaaS and you don't need a VPN. Cool.
However, the article refers to the enterprise side: if that service X that you're using runs on a cloud provider (AWS, GCP, Azure...) then they're probably see higher costs in areas such as mass storage or computing power and it would be cheaper to run their own infrastructure on a datacenter.
From your point of view as a use
Re: (Score:3)
When you have everything in house then you're going to wind up paying for hardware that sits mostly idle some of the time, or even some which sits completely idle most of the time.
When you have a cloudy architecture you don't have to pay for hardware you're not using before you use it, but instead you pay for someone else to make a profit off of your day to day computing needs. Someone still has to buy the hardware, and pay someone to install and maintain it, and eventually to upgrade it. And someone has to
Re: (Score:2)
Part of the problem is the "on cloud, you stand up a crappily secured project just like you would internally, but not it has a public IP address".
So while "ugh, the VPN iis inconvenient, it's so cool that the random instance I spin up is just plain available" is convenient for you, it exarcerbates the security picture.
If you are doing it right, you shouldn't have to fear public availability. By the same token if you are doing it wrong then you are not that well protected by a private network. However in pr
Re: (Score:2)
It does depend on scale I suppose. We did the study internally a few years ago (2018 I believe) for our HPC cluster. We accounted for buying the machine, paying for power, paying for the building, and salaries of a couple technicians, a couple engineers, a manager.
Our estimation was that to run our 100-node cluster with GPUs in EC2 and spend less money, we would need the machine to be less than 20% busy on average. We looked at SLURM logs and GPU utilization logs, CPU utilization logs. We stayed on premise.
Re:Cloud and remote work (Score:5, Insightful)
For an org that has there **** together cloud is far cheaper.
On prem requires excess capacity already spun up. On prem resources need far in advance planning. On prem lacks burst capacity capabilities for the most part. On prem also has an additional overhead of operations overheads to maintain, physical and the virtualisation platforms.
Now there are very valid reasons for on-prem kit. Lots actually. But for most organisations cloud solves huge volume of issues.
It falls apart when that org however forgets the cloud does require maintenance. On-prem forces orgs to not forget about some level of maintenance.
Re: (Score:2, Troll)
Trust. Compliance. True Cost. Down Time. (Score:5, Interesting)
Suits don't like engineers because of cash flow and lack of conformity. So they thought it was a good idea to rent their tech. Except it isn't as simple as it sounds. There are many things the local techs provide that have nothing to do with where the data is hosted yet, when they get fired because of cloudiness and the staff still have tech issues there is no one to talk to onsite. You get to talk to a tech support person on phone who may or may not be intelligible. I could go on. But I won't. Fuck the cloud. I never liked the idea, and I damn sure don't like the implementations.
Re: Trust. Compliance. True Cost. Down Time. (Score:5, Informative)
Re: (Score:2)
From my observations, you need _more_ people for cloud operations in a typical IT focussed enterprise, not less.
Re: (Score:2)
Yep, same observation here.
Good reasons (Score:5, Informative)
There are many good reasons for moving some things to the cloud. Services can be great. Prices low for low usage, no expensive licenses, and operations costs included.
Virtual server or 24/7 workloads has never been a good case.
Front-end servers exposed to the web makes sense as well. And so does things that are scaled very dynamically.
Train an AI model might be way cheaper than buying the hardware. That is like a dynamic load. No use 90% of the time.
Thus it is all about picking the right workloads to cloudify. We have many containers as part of web solutions that can scale. And lots of services.
Security, I would say it is better in the cloud. If you run services. Everything might be patched before the patch is publicly available. You have very fine grained access control. You have very good virtual networking and firewalls with tags that could allow those close to the app (DevOps /devsecops) to control the rules within tight restrictions.
Terraforming and other stuff is easy to do on-prem as well.
Re: (Score:2)
There are many good reasons for moving some things to the cloud. Services can be great. Prices low for low usage, no expensive licenses, and operations costs included. Virtual server or 24/7 workloads has never been a good case.
I think you missed a use case: because that's our vendors' strategy. I don't know if you can even buy on-prem Exchange any more. ServiceNow was always a cloud-only solution. Atlassian, GitHub, Grafana Labs, and virtually everyone else are making it very clear they develop cloud-first, on-prem maybe later.
What I'm curious about is whether those vendors will decide it's no fun running their own data center and/or buying one from Amazon and change strategy. We'll see. I personally don't care if we spend on lic
We've taken over a few failed cloud migrations (Score:5, Interesting)
We got an angry call from a company that wanted us to clean up the mess their previous IT company left. The lights were on, customers were coming in, but they couldn't work because a data center three states away was inaccessible due to a storm and their previous IT basically said there was nothing they could do. They didn't spring for the redundant hosting package so there's only one copy, which is now inaccessible, so sorry. We sold them a server. By the time it was ready, their datacenter was back online and we moved the data back.
We also cleaned up a company they that their previous IT told them they were too small for a server. 20 PCs, twice as many tablets, lets get rid of AD and move them to Office 365/SharePoint! No Azure though, it costs too much. Wait, how do we setup individual profiles across all the machines without AD? How do we deploy printers without GPOs? How do changing things like screen lockouts, sleep settings, Windows Updates? Well, they connected to each machine remotely and used scripts or did everything manually. After their first contract was up the previous IT crunched the numbers and found they were spending way too many man hours supporting this serverless setup and raised their prices, which wiped out all the savings the client had from getting rid of the server and then some. So, they came to us. We got them working right with their current setup, because the previous IT people were incompetent, and they are in line to get a server when their budget comes in this year.
Are their uses cases for it? Absolutely. None of our client's have onsite Exchange/email anymore. Email is a perfect thing to move to the cloud. Quite a few examples of Line of Business Software have done the same. That really does make maintenance easy for us. Oh, somethings not working? We'll call the vendor for you. Yup, they said its fucked and they are working on it. These vendors move their software to the cloud. We didn't sell it to them so its not our problem, but we still look helpful. I wonder how many companies selling their 'cloud' programs are going to move back to locally installed/hosted setups though? Every client we've had that moved to QuickBooks Online has regretted it. We have a few clients that need 3-4 of these different cloud programs to all be working for them to do their jobs and it seems they can't go a month without some outage somewhere causing them headaches. If they'd all been in house it wouldn't be an issue.
Re: (Score:2)
Agreed. The cloud is perfect for things like Exchange/email, DNS, simple web. SaaS and PaaS is where the cloud makes sense. If you are just lifting and shifting VMs to the cloud (IaaS) then you are not doing it right.
A lot of these places shifting back to on-prem will be complaining in another 5-7 years when it comes time to replace the hardware they are buying today unless they plan for it. Chances are they will not be ready.
Case studies for cloud vs. on-premise? (Score:2)
Re: (Score:2)
You can do local, on-prem cloud with Openstack and/or Kubernetes/Openshift. You can later deploy the app you developed to a few clouds that run Openstack or many of the Kubernetes/Openshift options.
Constant fire fighting, usually cost based... (Score:5, Insightful)
I work for a large corporate and the majority of our stuff is AWS cloud based.
Every developer needs to have at least some basic devOps skills, but we obviously have a dedicated devOps team.
A huge percentage of their time is maintaining a constant barrage of change, as old systems are replaced with new and once new systems become old and are themselves replaced, sometimes with as little as a few years inbetween.
Whilst that's fairly normal, it's the COST based fire fighting which is the most pertinent - it seems the team are forever in reports mode related to cost, trying to find rogue bits of infrastructure slurping up the cloud budget for no valid reason.
It's really easy to make cloud based services a dumpster fire, because you end up giving the keys to turn on new services to so many developers and make it super easy to do.
Like I said in comment intro, "Every developer needs to have at least some basic devOps skills" - I question that, I seriously question it, but that's where we're at with modern software dev.
With a microservices will solve everything mantra, clusters of pods spawn like bacteria, linked into all manner of poorly documented services.
We recently did an audit of our teams stack remit - it wasn't pretty viewing. Projects reaching back a decade, some which nobody in our team had ever worked on, with scores of s3 buckets, lamda functions, SQS services, Kubernetes spaghetti, ec2 instances lurking in the background, churing out logs nobody reads.
That audit was done using automated chart building software (and a lot of coding) - and the resulting diagram, if printed out to be readable, wouldn't span about 4 sheets of A0 paper. That's what a decade+ of constant change, including team change, does, when it's not managed.
Would this have happened if the bulk of compute power was maintained in-house?
Hell no, it then becomes super valuable - "sorry, we don't have budget to spin up a new server, do you really need this?"
"Why are you building a k8's cluster for a side project 5 people use?"
Cloud can be VERY expensive (Score:2, Interesting)
I know there are use cases, e.g. email, application level cloud (I'm thinking Sage accounting, Saleforce CRM) that make sense, but for general server provision in the context of bespoke software, cloud seems to be utterly uncompetitive.
In my case, developing complex financial analytics software, with a fairly heavy compute and data demand, I priced up cloud at about 25k per year (AWS Linux) for 1 server.
What I actually do is buy ex-data centre rack servers for about 1k a piece with similar specs to those 25
Re: (Score:2)
Cloud services are great for spike performance where you need a LOT of CPU time for a very limited amount of time, but you need that do be over quickly. It's far cheaper to buy CPU time from a cloud provider than to buy some beefy systems that idle away 99% of the time so you have that peak performance for that 1% of the time you need it.
Anything with a "standard" workload where you need constant but low/average performance gets really expensive really fast in a cloud environment and you probably want to do
Re: (Score:3)
Salesforce is frankly terrible. The only thing it offers significant advantage in is scalability. Even there it is kind of scary. You are required to tell them before you do a load test for example because it can impact other customers. Think about the ramifications of that for a second, a DDOS on some other Salesforce customer... yeah that's your problem now. It's easy enough to get your data out of the system, they offer various ways to export it, but they don't let you export your logic. And while they d
Re: (Score:2)
That's gonna be painful (Score:4, Informative)
Cloud services are a lot like a marriage. Getting in is fairly painless and easy, but if you want to get out, it gets messy, expensive and if you're not careful you'll find that half your stuff is missing.
The great life cycle of IT. (Score:5, Insightful)
When it was invented, compute was all centralized into large systems that predate todays mainframes and accessed remotely via compute sharing systems (aka time sharing).
Then the PC came along, and businesses repatriated their workloads to their own servers and the desktop. The justification? Security and cost.
Then the networking and Citrix came along. Companies started centralizing their workloads again, turning the PC into a dumb terminal that just accessed shared compute, again. The era of "thin clients". The justification? Cost.
Then companies started building more complex workloads in the desktop PCs and apps, because it cost too much to run. The Era of "thick clients". The justifications, again, mainly revolved around cost.
Then companies decided that thick clients were inefficient, and started pushing the idea of web based applications. This came along with centralizing workloads, again. This time it was called "cloud".
Now companies are moving back out to the client again. This time, they call it "edge computing". Again, the justification is claimed cost savings.
5-7 years from now the cycle will repeat itself again, with a new name for the same rubber band motion, always claiming cost savings.
Lift and Shift is the most expensive strategy (Score:4, Insightful)
As usual, the migration I'm part of is just a 'lift and shift' operation (meaning they want the cloud to run just like their data center). They are following the IaaS model.
On one hand, it makes sense for a world-wide organization to use the cloud for the availability and redundancy. On the other hand, just lift and shift is the easiest going to be the most expensive, (easy=expensive).. The real cost savings, if they are to be had, would be to refactor the databases, webservers, and so on to take advantage of the cloud infrastructure, depending on whether you want IaaS, PaaS, or SaaS. Only with SaaS could you get away with a minimal IT staff and IaaS would need much more staff. Maybe using the cloud native databases or serverless would work better and cost less overall than IaaS and your IT staff still running 'classic' databases and webservers. Or what about containers?
Lift and Shift might be a good way to start to keep your company running, but then there needs to be the effort (and budget) to refactor your code and data to take advantage of the cloud based on how you want to run it.
Like I said, not an expert but learning. Every time I read a story like this, I am left with a feeling of 'they' haven't really explored how to best use the cloud. The project I am currently on is the third one to go to AWS. In studying up on AWS they have many tools to track your usage and billing, and many tools to estimate your costs. And there are caveats on which storage you use and how you want to move your data around and what that will cost.
I would like to see some case studies by people who are familiar with what it takes to operate an on-premises set-up versus and properly configured cloud setup. Such a case study would hopefully give metrics on what it costs and how much effort to refactor existing code into a cloud project that maximizes efficiency/cost savings (for a true comparison, rather everyone jumping on the latest hotness and then jumping off six months later after their half-assed effort didn't work out)
Re: (Score:2)
This may be true, but it is:
a) Expensive to do that transition work
b) Doesn't necessarily improve the product you are offering to your clients (ie that work may be better invested elsewhere)
c) Further locks you into Cloud
Sticking with "lift and shift" and using cloud versions of "commodity" services (eg RDS for Postgres, or Elasticach
Former IT guy turned sailor (Score:2, Insightful)
I thought the cloud was a dumb idea back when I was doing IT. Why would you want to be reliant on 2 companies, your ISP and your cloud service provider, to do business? Why would you want to give big data access to your proprietary and customer data? Why would you want to get stuck paying whatever they asked, when they started raising prices on you down the road?
If you were still my customer, you'd be happy knowing that while your friends are paying big bucks to migrate off the cloud, your data never lef
Devs believed cloud-magic would fix their issues (Score:5, Insightful)
The Devs were I worked believed the cloud would fix all of their issues, as they blamed all of the issues on internal infrastructure. They believed in non-existent magic and oversold the cloud fixing all issues. They spend large amounts of time to "fix" the app by re-writing it and ended up with much the same as the app they started with--ie still broken but in different ways (it was a giant lost-time-accident costing 1000's of man-years).
They believed the internal network was the cause of their issues. In reality, the cause of their "network" issues were all in their app (getting backlogged and not answering, and leaking sockets cause denial-of-service in inbound/outbound network sockets). Their CPU and RAM issues were also generally bad code, and the cloud kind of fixes the leaking RAM issues because the cloud instances gets restart/rebooted often enough.
And once you move to the cloud and have the same issues, but it costs more you get pushed to move back.
I have worked with a lot of devs over the years, and less than 1-2 in 20 actually know what they are doing. But all 20 are supremely confident they are experts.
"The Cloud" is just a remote server you don't own. (Score:3)
Keep your main data at home, run everything from there, and keep a backup on a somewhat remote server in case the building catches fire.
Perspective of a small business (Score:3)
It's interesting to hear about all the alphabet soup of services that big companies use on AWS and elsewhere. FWIW, I ran the infrastructure for a mom'n'pop business, and I put their website onto an EC2 instance. This saved worrying about uptime and connectivity, and since the website was fairly low traffic (hundreds of pages/week), the instance could be small. I don't recall the exact cost, but it was low 2-digits per month, so low 3-digits per year. Definitely worth the cost.
Note: This was a few years ago; for all I know, costs may have gone up.
The only problem I had with AWS was the inability to set a limit on the monthly charges. Sure, you can set alarms, but you can't tell AWS that costs are never allowed to exceed X. Which means that I always worried a bit about the account being hacked. Obviously, you can do a lot to enhance security, but there is no 100% guarantee...
tl;dr: Cloud can make sense, even for SMEs, but you have to know exactly what you want.
Re: (Score:2)
Technically you should be able to put a budget action to turn off services, but I have never seen it in action before. https://docs.aws.amazon.com/co... [amazon.com]
Leaving the cloud... (Score:2)
should be called "rain". Why make up a name when we have a perfectly good one in wide use?
Hype - You think "the Cloud" is disappointing? (Score:2)
I think many of the people here saw this coming. Moving data onto someone else's iron is not cheaper, and maintaining it is more expensive and subject to glitching and outages. There's no substitute for physical access, despite all the provider assurances. There's also no substitute for being 50 ft from the kill switch. There are use-cases for cloud services, but they are not as common as hype was indicating. Very few businesses benefit from a remote deployment, and nearly everyone here saw that coming.
Turn
The Cloud Is Just Someone Else's Computer (Score:2)
Perhaps 1990-2010s-era server rooms will be back, but hopefully with better monitoring and remote assistance, because any sysadmin will tell you, the budget to get everything you'd want to run the system
Yet cloud providers continue to grow (Score:2)
AWS 4Q 2023 revenue grew 13% YoY to $24.2 billion. So more workloads are going on the cloud than leaving.
https://www.cnbc.com/2024/02/0... [cnbc.com]
Where cloud has failed in my experience (Score:2)