Google, AWS say it's too hard for customers to use Linux to swerve Azure When moving to the cloud, companies with significant investments in Microsoft infrastructure wares simply can't afford to rewrite everything for Linux, so they end up migrating to Azure to dodge the markups Redmond charges for running its server software in competitors' clouds. Or so say both Amazon and Google in their latest …
"...it would take years and years to modernise to Linux after migrating as they would essentially need to rewrite all the Microsoft-based applications that they have accumulated over the years which is very challenging for most enterprises."
No, really??!! I would have NEVER guessed. I mean, after all, I've only been trying to tell people here [on this forum] this for *years* (in regards to Linux desktop) but everyone wants to live in their own little fantasy world.
Next up: Sky appears blue.
But only on weekends :p
Certainly anything written non-portably for Windows is going to have to be rewritten to run anywhere else. That's a fact of life i moving platforms. It's a very long time since I wrote anything for Windows but my experience over a few years was that it was just a perverse system to write for. In reality it might be easier than indicated here* to move stuff based on DotNet over to Linux where is is supported with official Microsoft implementations.
OTOH Pinta, which is based on DotNet, is distinctly crashy here. Admittedly I'm asking it to handle a really large map with multiple layers but Gimp doesn't mind the same material (I use both - some operations are easier on Pinta, others on Gimp, at least for a non-graphics person who's officially been declared by his wife to be "not artistic").
* Note that this isn't really about porting S/W, it's non-Microsoft cloud vendors building a case against Microsoft about pricing differentials so Rice-Davies applies. AIUI Linux is actually a major client OS on Azure.
I think perverse is stretching things. There are times when an irritating number of APIs have to be called in Just The Right Way (I seem to remember fiddling with ACLs a long time ago fell into this category), whereas some other operating systems just used 'bung some values into this struct and pass it to a function, job done' but on the whole Windows' documentation is good, and the API coverage extensive.
It works both ways though, there are so many areas where Linux historically and BSD (still, right now) are lacking in documentation or the solution is 'run this sysctl, parse the text output' or 'this isn't a user mode API, even though there's a good reason to use it in user mode'. Whereas for Windows the API has existed for years, has better documentation, doesn't use fragile text parsing, and covers more use cases.
I've also watched more than one presentation where Linux and BSD enthusiasts compare the Windows way of doing things with the Unix way, (for more modern technology such as USB IIRC), and the Unix way is incredibly painful.
Windows dev tools are, in general, much better too. I tried CLion under Unix and its obsession with CMake made it unusable for my needs. I had to resort to Qtcreator, which involved faffing around with config files. It's not bad for free, but Visual Studio has handled all the scenarios I wanted and only vanishingly rarely needing to drop into an editor to tweak things.
I'd counter in that most software is written in a higher level language and much more easy to port to another OS.
The biggest gotchas I've experienced uplifting software are inconsistent paths on a case sensitive file system. And properly serializing dates and times with an offset to gmt/it.
I think getting out of MS SQL is much harder for more software than getting out of Windows. But there are Linux versions of that.
I think getting out of MS SQL is much harder for more software than getting out of Windows. But there are Linux versions of that.
Linux versions maybe, but MS still wants the licence money for it, and the problem described here is MS making it hard for people to use Cloud other than Azure using licence money as the weapon.
If you already have Windows SQL Server licenses for 2017, 2019 or 2022 then those licenses can be used against Linux SQL Server installs. You dont need to buy new ones
We run our HR stuff on SQL Server on Linux as its a far cheaper option than the other supported alternative, Oracle
I dont see how a company wanting to charge license fees for its own products is "making it hard for people to use Cloud other than Azure using licence money as the weapon"
The 1970s are over, and K&R are dead (or at least I hope). The language today is less important than the frameworks and libraris a softare is build upon. While in the stone age of punched cards most ot the software was on the card themselves, but a few OS services (more or less, like DOS...), today an application, even a sever one, may rely on a large number of specific libraries. Porting it to a different OS without those libraries is not easy at all, even if the the language do exist on that OS. You would need to rewrite and rertest a lot of code.
And if you had developed an applicating relying on specific Windows APIs and services, good luck in porting it to Linux easily. You can try with WINE, but if it is a critical application for you company you paid for, and not some software you download illegaly from somewhere and you want to run it without a Windows license, you're not going that route....
Moreover Windows do ensure far better backward compatibility than any Linux distro (even RHEL) - which makes it even more important for applications you can't recomplile at your will.
It’s simpler than that - it’s discriminatory pricing, from a supplier with a dominant market presence.
Not so much on the Cloud - though noted it is #2 - but in Windows as the OS and SQL, Exchange, Office as the standard business fit software. Consolidating Faf client Office into subscription M365 doesn’t help either.
It is perfectly normal for any vendor to give themselves "a discount" on their own damned products! Just because Windows doesn't come for "free" like open source software does is not Microsoft's "fault"; they don't "owe" companies like Google or Amazon anything. Nor is this an issue of "monopolistic behaviour." You don't need a monopoly to take advantage of locking in your customers; just ask Apple about that.
I think you need to get on message, if you want to benefit from the AI gravy train. By the time businesses encounter the limitations of LLMs and AI, their sunk costs will simply encourage them to throw even more money at you… with a bit of luck you may get to a well funded retirement without actually delivering a working system.
how many google and amazon software/services can you run in/from Azure at all? (NOTE: doesn't mean making remote calls to services, means running their software IN Azure). I wouldn't count Amazon's linux distribution(or other re-badged that was otherwise open source under another name) as there are many free sources of Linux systems. Talking ground up written by Google/Amazon.
I don't have skin in this game I don't like any of the IaaS clouds, I'm all on prem (even for my own personal dns/web/email/file hosting I run my own systems in a colo).
I suspect other big software companies that provide cloud services are similar, guessing Oracle provides added discounts for running Oracle products in Oracle cloud(or otherwise hosted by Oracle). Probably others too(Salesforce? SAP? IBM?)
That's not the issue. The problem is this: you have a Windows server on-prem, and a license for it. If you move it to Azure then there's no license cost because Microsoft converts it to a Cloud License. If you move it to AWS or GCP then you need to buy a new license because Microsoft doesn't allow you to convert it to a cloud license. That's why everyone moves their Windows workloads to Azure.
"you have a Windows server ... a license... If you move it to Azure then there's no license cost because Microsoft converts it to a Cloud License. If you move it to AWS or GCP then you need to buy a new license because Microsoft doesn't allow you to convert it to a cloud license."
Leading a charmed life I have never had undoubtedly pleasure of dealing with Microsoft licensing or much to do with MS generally but reading the above I would guess that if the putative client's Windows server (license?) were transferred to an unshared bare metal server in the service provider's facility then a replacement cloud license shouldn't be required?
I could see a pool of physical servers being multiplexed between network connected storage to provide a single bare metal server+single owner+single OS as required by the licensing.
Even in the 1990s there were those among us who from bitter experience were warning against getting into bed with Microsoft as you not only guaranteed to get the usual but also something far less pleasant and utterly incurable.
“ The problem is this: you have a Windows server on-prem, and a license for it.”
I would say the the main problem is that companies even consider this scenario. Even *without* licenses a lift and shift to IaaS is hugely expensive compared to just stacking just putting their own iron in redundant datacenters, colo if needed.
Even without own hardware, I ran the numbers for a couple of tomcat servers. On average MSFT, AMZN and GOOG were triple the yearly cost compared to a local hoster (that offers 24/7 support & monitoring). YMMV, since all of them have murky pricing even for what only laughingly can be refered to as “PaaS”.
A pox on all their houses. I have to use all of them, but I sure as hell don’t like it.
Depending on scale, that can be true, but I've frequently experienced the opposite. Colos near me have relatively high base prices for having a single server in them. Even ignoring the cost of the server and costs to manually go repair it, just having the space there could pay for several VMs. How cost-efficient the approach is tends to depend on how powerful a server you would install in the colo if you could. If it's something massively expensive, then buying it and renting space to store it will probably be less expensive than renting the equivalent VM. Many small businesses or projects don't have that requirement and can work just fine on smaller VMs. When I've done the numbers, it often ends up being more efficient, sometimes substantially so, to rent those from some cloud provider rather than obtain a basic server and find a place it can operate.
Upvoted , and to add - So a Microsoft app needs to be re-factored to run in a non-Microsoft cloud? So - what?
Some Amazon services and GCP services are unique to those platforms as well. so what? Portability between those clouds also required *gasp* refactoring!
"Refactoring is too hard" is only true until you actually scope the work. If your application is so complex that it is a MAJOR or impossible effort - then you were stupid to put it on someone else's infrastructure in the first place, the application is crap, and probably should be re-written, not re-factored. Suck it up, Buttercup...and get coding.
It's not like there is some pricing advantage to going to something else. Anyone who thinks they are going to run on AWS to save money is dangerously misinformed. Same for guarantees of uptime. Why move your thing to some other thing that was not built to run your thing?
Azure has a functionality that AWS or GCS cannot match, same as the other way around.
And the cloud is still someone else's computer that you pay too much to use.
"Azure has a functionality that AWS or GCS cannot match, same as the other way around."
No, it doesn't. It's not a functionality. It's an artificial licensing restriction which is an abuse of monopoly.
If you decide to use a proprietary AWS or GCP service that has no easy exit path to another provider, that's a decision you're making with your eyes wide open. Most sensible people don't box themselves into a corner when architecting cloud-native by using services for which there isn't an almost drop-in equivalent on other platforms.
If you've been running Windows servers on prem, from the era of dedicated physical tin, then moved to virtual machines, there's no technical reason at all why you can't move those servers to AWS, or GCP, or Azure, or Oracle Cloud, or any other cloud that provides IaaS services.
If Microsoft had a rule that said you can't move on prem licenses to the cloud full stop, that would be annoying, but fine.
It's the fact they treat Azure as a special case that is blatantly an illegal abuse of a monopoly in one sector being used to gain an illegal advantage in another sector.
It's not rocket science, and they deserve a massive fine, an immediate cease-and-desist, to pay compensation to customers that have overpaid to use Microsoft products on other cloud vendors, and to waive egress charges for anyone that was affected who subsequently decides to abandon Azure to migrate to another cloud vendor.
The whole *point* of many commercial products is to tie you into their products and provide a steady revenue scheme, it is not in their interest to make it easy to migrate away.
'Linux' is no different, unless they actually mean 'open source that works, with decent documentation'. A not inconsiderable amount of 'Linux' software doesn't work particularly well, the documentation is appalling, and to get it to work properly or easily needs proprietary components or paid consultancy which the software provider *oh so surprisingly* also sells. Funny that the documentation never becomes amazing whilst consultancy is on offer, isn't it?
Take Kea. It's actually pretty decent. It's (mostly) open source. The documentation is good. It's pretty easy to get running. There aren't too many unexpected gotchas (which is more than you can say for BIND). Yet the real enterprisey auditing hooks are paid for (to be fair it's a fairly specific set of hooks, most of the functionality you'd want is gratis). This is probably as good as you're going to get, and the (paid) support is entirely optional.
Pot, kettle for AWS. In theory I understand you can run AWS apps locally, I vaguely checked it out after looking at the requirements for the supposed situation when Elite Dangerous stops being sold and the AWS based server software is released 'for free'. In practice the list of dependencies and the effort to get it running locally appears far beyond the informed enthusiast level, and into the realm of paid professional.
By definition software or services that sell for money will generally work (to some extent), and have advantages over free options (usually), otherwise no-one buys it and the product dies. Companies get a leg up over designing a product from scratch using generic completely open source components with no appreciate commercial influence.
It's not surprising Microsoft did this, either. Back before the turn of the century SQL Server licensing (7, I think) did not prohibit a one user model for SQL clients on servers, which could be used to multiplex websites and save large amounts of money on licensing. Microsoft changed the online licensing whilst the product was still on sale and as the Internet grew in popularity, despite the earlier (printed) licensing contradicting this. SQL Server 2000 had specific licensing designed around webservers right from the start. Mind you, 7 wasn't one of the better products, and the management interface was slow and limited.
Almost certainly. The question is what flavour or strength of poison you wish to take and the associated technical debt. Unless you're lucky it's a lot of effort to bring up a product from scratch with no external dependencies in a timely manner.
The worst dependencies I've seen in the product I've dealt with heavily for years have mostly been third party programming components or in one case supplied Microsoft components (SQL Reporting Services had some notable capacity limitations in earlier versions, now fixed). The time to market and cost for an in house developed alternative to the third party components simply weren't acceptable - the later pain and technical debt that needed to be dealt with were ultimately worth it although I didn't enjoy the workarounds at the time!
For non programming components or the languages/runtimes themselves the lock in is acceptable, particularly when most other customers are using the same platform and languages. Microsoft are in general fantastic at backwards compatibility unless you're daft enough to pursue some of their new tech immediately and get burned by some of their desktop forms, web, or mobile ventures that they dropped in a relatively short time frame. Thank goodness we never bothered with technology such as Silverlight; others did and the rewrite was expensive.
'once you pay the Danegeld you never get rid of the Dane'.
Retaining 'freedom' is costly, and requires integrity and vision, along with money, when it comes to running a software stack that is effectively 'ours' from top to bottom.
Paying someone else to control what is precious to you, sooner or later is going to be a cause of regret, and by then it may well be too late. Suck it up, you've made your bed, now you get to lie in it.
MicroSloth - Always Be Misstepping.
Charging four times as much to run your own software on a competitor's hardware does sound a little steep. Maybe a nice 10% markup would be enough? It's not like they don't already have an abundance of money. Where are those WWII profiteering laws when we need them?
Still, this isn't about M$ overcharging. It's about Gezos and Bezos trying to get (and keep) more customers. Seems they have some resources; figure it out, folks!
Bits is bits, and data is data. The customer doesn't need to know how the interface works under the hood, so long as they can operate it. How many of us can build our own cars, vs how many of us know how to drive?
Where G and B could really shine is to do a little work themselves, figure a way to have the M$ customer get their same reports running on Ubuntu or whatever - even ugly-fying the output to make it *look* like something Redmond would do - and then offer that API to customers at a fairly discounted rate to get them in through barn door, and stay there, content if not happy.
For extra credit, they could make it look like CoPilot was helping.
As always, this issue only effects the small guys! If you have an EA with Software Assurance you can port your licenses into AWS, Google, et. all without issues. This is no different than moving licenses to VMware! You needed Software Assurance! The fact that most people didn't buy SA when they virtualized and Microsoft didn't sue them is beside the point!
The cloud providers are not going to allow you to get around that requirement, so you either rent the license or you buy SA.
For some things renting makes sense (for us). For SQL server licenses, if there is an opportunity to downsize, we can and are not stuck with a 3-year commitment for expensive per core licenses we no longer need. It's the same for things like RDS!
Where this issue does come into play is Windows Server. As of 2019 you cannot bring your SA Server licenses to AWS, etc. for shared instances. you have to rent a license. Yes, Windows instances are more expensive than Linux, that's no different than on premise! That's no different than it's always been! It's a business decision, is the cloud more cost effective than maintaining an on-premise data center for your business? Yes, you bought that license, yes, you cannot use it in the cloud, yes that sucks. Do you really need that 16 core file server in the cloud? No, you don't, you didn't need it on-prem either, that's why you virtualized! You had to buy licenses for all those virtual servers (including SA if you were honest). So now a 2 core 8 GB file server costs you $75 /mo. (T3a.large in AWS), even with your license, an equivalent server in Azure will cost you more! No matter what MS tells their customers, Azure ends up costing more even with bringing your own licenses, with poorer service! PERIOD!
For most SMBs moving away from Windows is not an option, they didn't build their apps, they bought them, and they run on Windows!
"For most SMBs moving away from Windows is not an option, they didn't build their apps, they bought them, and they run on Windows!"
Wrong. You can run most of them with wine, under a secured, encrypted Linux. Being doing that for ages, all sorts of clients.
That’s actually wrong cliffwilliams45. There is no Software Assurance on Windows Server, no matter what Volume Licensing program the customer uses.
Also, there are critical products, like MS Office, that Microsoft use licensing terms to block completely from being deployed to GCP etc. This is a real issue for many enterprise systems, like SAP etc, that require a copy of office running on the backend servers to output reports in Excel/Word. This is a real issue that barely gets a mention.
>” This is a real issue for many enterprise systems, like SAP etc, that require a copy of office running on the backend servers to output reports in Excel/Word.”
There are tools other than the MS Office suite and with OpenDocumemt standards things should be less problematic than they were in circa 2002 when I first had to grapple with this problem.
Although, as you note ( in another subsequent comment, MS have done much to make customers wary of non-MS solutions…
The fact you are raising the point that this is still a problem ( and potentially an even bigger problem with cloud) shows how little MS have actually progressed in producing software that actually helps users and businesses.
Our two Azure VMs exist for Great Plains/Dynamics. I am on good terms with the comptroller, but if I suggested that we migrate off Great Plains to save on licensing costs, I might not survive the discussion.
The commentariat is free to tell me that this is our fault for not rolling our own AP system with GPL software.
Microsoft are the masters of licensing. I'd shudder to think of the amount of licensing paid by US DoD. Shudder. Then the Secure Technical Implementation Guide (STIG) must be applied to secure the platform for interoperability by Tier 1. Now nothing works correctly for the end users Tier 3. Check out how purdy Sharepoint works after whacking it's scripting languages. Don't even get me started on Certficate Authorities (CA) and Realms. That's on Azure. Wait! There is more! Got to use the AWS. Need those licenses there too. You really want to the golden chained belly dancer, look at the latest healthcare implementation. Yeah baby it's purple and its virtual. But hey got to feed those hungry hungry Military contractors. Rant port 6533 closed.
It's really not that hard to port MS apps for Linux. Most .Net apps can be updated to the latest .Net without too much issue. At least web apps.
SQL server can range from easy to nearly impossible though. Depends on how it's used.
I've spent about half my time in the past decade shifting C# apps to Linux and containerization. It's not so bad.
DO NOT TRUST AMERICANS
Ignoring Emperor T and his henchMusk, this is just scare tactics by a bunch of greedy americans who really like their profits. So, if your board like feeding beeelions to the greedy untrustworthy yanks, feel free to keep the status quo.
However if you actually value your data and systems, and want to keep more of the profits, why not hire some genuinelty decent developers, sysadmins etc. and (oh the horror) pay them what they are worth, and keep them current with relevant timely training. Long term, it will be money well spent.
If your critical business apps are that reliant on Microsoft, please continue paying your Microsoft tax (+/- whatever they deem appropriate in terms of future pricing changes) in perpetuity.
There will come a point where you'll realise that it would have been cheaper to just make your software platform agnostic in the first place.
But if you're paying that tax happily each, then you'll never realise it until it's too late.
I can't understand any large company putting all their eggs in the Microsoft basket, especially with the push to Azure / cloud desktops now. You literally can run your platform on ANYTHING and you decide to lock into Microsoft? Then pay your stupidity tax.
“Google said that for companies that have built a dependency on Windows Server and/or SQL Server, it would take years and years to modernise to Linux after migrating as they would essentially need to rewrite all the Microsoft-based applications that they have accumulated over the years which is very challenging for most enterprises."
Modernise to Linux? Windows "accumulated over the years".
Microsoft must be reading that and hurting. Feels to me like they're essentially calling Windows a dinosaur. Next they'll be talking about "vendor lock in", just like people used to do about mainframes.
Remember Windows '95? It was rubbish. But right after its release, world + dog was writing applications for it. Not just large companies who bet on the market going that way (and who saw what happened to those who had bet on MS's earlier promises about OS/2, sunk their R&D budgets into it and then had the carpet yanked out from under them) but also smaller developers who had until then made little or no DOS or Windows 3.x application software.
Why? Because MS released a slew of cheap (sometimes even free) and easily accessible development tools, frameworks and suites along with Windows '95. Suddenly making applications for Windows '95 looked a lot easier, cheaper and more attractive. The flipside of that was of course that by the time Windows '95 became Windows '98 the application development market was firmly locked into the MS Windows ecosystem with vast piles of code that was less portable than the Rocky Mountains.
Which, of course, was exactly what MS intended. Cheap and accessible development platforms to produce code that was nigh impossible to port to any other OS, ever, was the whole point.
And the rest is history. For three decades it's been a matter of rinse and repeat.
It is possible to put Windows applications inside containers...
Also, Wine is a thing.
Plus I think they may be overstating the importance of some of the dependencies. MS SQL is, after all, an SQL database just like many others (some free), with differences fairly well understood. Web pages are still just web pages, no matter what server dishes them up.
Additionally, it's not exactly like AWS isn't engaged in creating vendor lock-in. How portable is Lambda? Not very? There's been plenty of articles about developments that have gone down the AWS Lambda route, got running easily enough only then to find that the scale-up costs are no where near as affordable as they'd hoped...
Having said all that, Windows is a thing and if I'd got legacy on-prem infrastructure running Windows and Windows-based applications, re-writing them would not be high up my list. Nor would be ditching all the on-prem infrastructure. The wisdom these days seems to be that running some on-prem infrastructure - a hybrid approach to Cloud - seems like a good idea. Windows on-prem infrastructure brings all sorts of useful things with it that are pretty hard to replicate (e.g. AD) on Linux. Linux on-prem infrastructure is messier to admin whilst providing your average corporate network experience and endpoint control.
Azure does seem to offer quite good options for hybrid use of cloud. It might be this aspect that makes people adopt Azure?
The point that has been missed in all these comments is that 70% of business application run on Windows Server. And the vast, and I mean vast, majority of these application are Commercial Off the Shelf (COTS) products written by 3rd party ISVs. Everything from SAP running entire global enterprises to small line of business systems undertaking a specific function at an SMB.
Enterprises buy these application from the ISV. They don’t write them themselves. So they can’t ‘port’ the application off Windows as they have no access to the source code. They can only run the code the ISV creates on Operating Systems that the ISV dictates (and is prepared to support).
And Microsoft have spent many many years getting the ISV ecosystem to almost exclusively target Windows Server (and SQL Server). I should know, as I spent many many years at Microsoft doing exactly that….
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
