Linux, not Microsoft, the real winner of Windows Server on ARM Microsoft now runs a bunch of Windows servers on ARM processors. Apparently, these ARM chips are quite good at their jobs and Microsoft might try converting entire categories of workloads over. All around the world the tech press has speculated on whether or not Windows on ARM will be showing up in on-premises datacenters. In …
Lol at dreams about SQL on Linux in someway meaning the demise of Windows Server even though Windows Server sales are still climbing!
We already know it's a cutdown version to keep certain devs happy. Production workloads that need any enterprise features are going to need to run on Windows Server.
"Or MySQL, Oracle, DB2 etc. on your choice of one *nix or another...."
I meant Microsoft SQL Server workloads obviously.
nb - MySQL and Oracle run just fine on Windows Server. Can't say I have seen an install of DB2 in living memory to know. And Oracle at least is much easier to install / initially configure on Windows Server than on *nix...
I use SQL every day at work. I don't really care whether the server is running Windows or Linux, just whether I can connect and run the queries I need and tune them.
To the extent that Windows is essentially finished, especially for RDBMS servers, this is the beginning of the end. That SQL is one of the "big dogs" of the database world, and that it exclusively ran on Windows was a big USP for Windows Server. It isn't any more.
You don't need Windows domain servers - AD in the cloud will do. You don't need Windows Server to run a DHCP server, a DNS server, a mail server (Office 365, remember), and now you don't need it for on-premises SQL (you didn't really need it for cloud anyway, did you?)
With the advent of Visual Studio Code for Linux, you quite possibly don't need the development stack I use every day, either.
You can only swim against the tide so long. The day of Microsoft dominance is ending. They are slowly, but surely, going back to being just one among the many, and the demise of Windows Server is simply another nail in that coffin. I suspect Windows desktop to be around a bit longer, but macOS, Chrome OS and desktop Android machines need to start selling in bigger numbers before that finally falls off the perch.
"Or MySQL, Oracle, DB2 etc. on your choice of one *nix or another...."
If your customers program uses MS SQL Server then that's what you need. If the Linux version lacks features that the customer's program uses then you're stuck with Windows server.
I'm interested in the Linux MS SQL Server for replacing SQL Server Express which gets installed for free when you install a program which requires a database.
Or ........converted to PostgreSQL/(commercial) EnterpriseDB for best performance, greater reliability and significantly better security.
An informal survey of enterprise Linux developers and companies has shown that only a few who are "contracted" for some Windows development are using SQLServer on Linux.
Furthermore, Linux SQLServer does not run natively - actually horribly, and is a total dud on Linux.
Let me boil the writer's argument down to one sentence: running Windows on ARM means that all servers will run Linux on ARM with non-Microsoft SERVER applications because Microsoft made such a mess of CLIENT applications Windows 8 for ARM?
MS are in a mess over all sorts of things, but I think they have a clearer view of what's going on than this author.
Windows Server on ARM is currently proof of concept for Microsoft. There has been no indication that it will ever see general release and Microsoft's strategy is clearly to get as many customers to stop running their own servers and use Azure instead. x86 support for things like Exchange is already available but Linux is where the growth is and already popular on Azure and even better suited to ARM than Windows.
I think what Trev is saying is that Microsoft's adoption will drive open standardisation that unintentionally benefits its competitors, who will further benefit from Microsoft's litany of recent failures.
Microsoft has always been hideously inept at nearly everything, but managed to maintain its monopoly by the circular dependency of proprietary standards, and legacy software locked into those standards. However, it wasn't prepared for this mass migration to new architectures, and now it's playing catchup, and losing, because it has little choice but to abandon its own legacy, and thus break the chain of dependency that kept an entire industry tied to Microsoft for decades.
Microsoft is still a big player, but it will become increasingly irrelevant over the coming years, and is inevitably destined for obscurity, if not extinction.
Personally I won't miss it, in fact I suspect most people won't even notice, as we head into an ever more abstracted Cloud environment.
Like Apple, Microsoft has the image of selling stuff that works. Even when it don't and they are just emulating others. Its all about the image and what decision makers (aka sheep) follows.
Just look at the uptake for Azure, As a techo you don't trust it and understand the complexities they must overcome before they can be AWS, but because your desktops and servers run windows, it must be the direction to go according to the decision makers.
Like a bad rash, Microsoft will be back.
I'm always confused by this kind of argument...as if people live on a completely separate planet to the one I live on. Microsoft positioned itself some time ago as the convincing number 2 in cloud computing, and has sustained increasing cloud market share over several years (i.e., they are gradually playing catch-up with AWS market share, though there is still some considerable way to go). Anyone who has had dealings with the company over the last few years (all of this current decade) knows that Microsoft long-since adopted a rigorous cloud-first policy internally. The issue for them is not how to maintain x86 legacy apps in the cloud - something that can be done technically, but gets little explicit attention from them. Azure, at the heart of Microsoft's ambitions, is very clearly not about maintaining some old x86 and desktop legacy! I don't get how anyone could seriously think that is the case.
My world is taken up with designing and implementing a modern microservice architecture for a continent-wide industry sector application on what is arguably the world's most advanced generally-available hyper-scale/high-availability technology platform, Azure Service Fabric, which is significantly more advanced technically than, say, Kubernetes (to which Microsoft contribute) or Docker Swarm (which Microsoft is committed to supporting). We are using the Windows version, but they ported Service Fabric to Linux because, very simply, all that counts in the cloud is consumption of CPU cycles and storage. x86 legacy? I think not. Service Fabric is in its fifth-generation and the foundation on which so much of Azure is built.
We all know the world has moved on since the PC-centric days of the 1990s. Microsoft has lost several battles since those days, to be sure. But the underlying assumption in the article and several comments here is that they have failed to grasp that the world has moved on. This is clearly untrue. It has been very clearly untrue for a very long time. It doesn't begin to describe their strategies and practices in recent years. It's really is high time people moved on in their thinking about market players like Microsoft. No point being stuck in the 1990s, now is there?
Yeh, the author sounds like a kid, or even a kid 20 years back going on and on how Java is going to cure platform awareness. Besides, when you think about powerhouse data centers, do you think Javascript ("web native")? No, you think about embeding Adobe Flash script in a jvm!
That is not how I read his argument. I thought his argument was basically that if the shift to ARM hardware is important enough for Microsoft to make the jump, then they are already losing the advantage of legacy application lock in, so they won't have any means to force themselves into a dominant position on ARM hardware. He believes that they will have trouble competing on a more level playing field.
Microsoft's choices regarding Windows 10 have ruined the brand name, and there's no need to carry any baggage or expectations forward to this new platform.
I guess that Microsoft will now join Apple (and no doubt a few others) in NOT inviting representatives of this site to Press junkets. How dare anyone criticise Windows 10, the Windows to end all other Windows for everyone on the planet? /s /s /s
The acceptance of ARM as a server platform rests no with MS but with all the application vendors out there. If they don't port the likes of SAP, WebSphere, Oracle DB and the multitude of other applications that businesses use to any ARM platform then it won't matter a jot if that platform runs Linux or Windows or something else.
Intel should be very worried about X86 but its day might be gone. At least (AFAIK) they have an ARM license but I would sense that inside Intel that there is a huge feeling of NIH , DejaVu and 'if only' all mixed up together.
now only if some of this lovely ARM Server H/W was available to us mere mortals at prices that were compatible to i5/i7 systems and off the shelf. That may be sometime coming though.
now only if some of this lovely ARM Server H/W was available to us mere mortals at prices that were compatible to i5/i7 systems and off the shelf. That may be sometime coming though.
Quite. I'd love to get my hands on few Cavium ThunderX2 boxes/boards to replace some aging x86 stuff with. At the moment the only realistic options for reducing power consumption are either crappy Atom (for some very light non-memory intensive workloads) or low power i5/i7 (or perhaps Naples, depending what AMD comes up with). Given a choice I'd opt for the Cavium over the others, given the goodies in the SoC.
This is huge. SBSA is the real threat to Intel.
It is huge.
It's also something Microsoft could have defined back in 2008 when they first acquired their own ARM foundry license. Had they done that (they even demonstrated Win 7 + Office on ARM) then instead of doing their level best to Utterly Ruin Windows by trying to be cool and down with the mobile kids by pushing Metro, Windows 8, etc (something that they continue to do to this day, plus they've added snooping into 10), we'd now be used to ARM servers and desktops, MS would still be top of the mountain, and we'd probably be happier with Windows too. Instead were seriously wondering about not bothering with anything MS at all.
Cock up.
This is definitely bad news for Intel, and by extension all current users of X86. That includes the supercomputer guys. Anyone who actually needs all that compute offered by Intel's biggest chips will be finding their lives becoming expensive. ARM is fine for what it's intended for, but a fire-breathing high performance general purpose CPU suitable for weather forecasts it is not.
Interestingly Fujitsu /RIKEN are contemplating ARM plus their own specialist extensions to make their next super computer. Expensive.
ARM is fine for what it's intended for, but a fire-breathing high performance general purpose CPU suitable for weather forecasts it is not.
HPC teams already use heterogenous hardware mixing x86 with GPUs because x86 hardly shines at parallel vector work. While x86 delivers great single-thread performance that necessarily the most important part of HPC. ARM chips already come with optional hardware acceleration packages, throw in FPGAs and GPUs and, at the right price*, the HPC crowd will be drooling.
* If you look at some the biggest HPC installs it's obvious that purchase price is not that important. Increasingly, it's important to have something that doesn't need its own dedicated power station.
I do not get this part "Anyone who actually needs all that compute offered by Intel's biggest chips will be finding their lives becoming expensive. ARM is fine for what it's intended for, but a fire-breathing high performance general purpose CPU suitable for weather forecasts it is not."
It is enough for ARM to be competitive on generic MT-friendly loads. On the other front you have competition from OpenPOWER, led by IBM. These two taken together, with Linux as a platform, are enough to provide some competition in the datacenters. Not much of it, but some is surely better than nothing. And who knows, Microsoft could port Windows to POWER architecture if they wanted to, that won't be very difficult (I know, I've seen the sources).
HPC teams already use heterogenous hardware mixing x86 with GPUs because x86 hardly shines at parallel vector work.
Well, it depends on the workload. Xeon Phi is quite a big beast, and we'll suited to some workloads. As ever, it depends.
GPUs are problematic for some workloads. Their downfall is latency; they're (still) all about loading up some data, doing a lot of math very quickly, and then unloading the results. For some problems this is less than ideal. Machines like RIKEN's K is very impressive because they did so much to reduce data sharing latency in the machine, which gave it an unparalleled peak:mean performance ratio.
ARM chips already come with optional hardware acceleration packages, throw in FPGAs and GPUs and, at the right price*, the HPC crowd will be drooling.
Drooling, but facing a massive code rewrite!
If you look at some the biggest HPC installs it's obvious that purchase price is not that important.
That's mostly because the chips they use are the same (more or less) as gamers / server farms use.
It costs Intel around about $6billion to do a step in their design, and it's about the same for everyone else doing circuits that complex and fast (be it GPU, Ethernet switch, whatever). If Intel stops bothering, or if NVidia give up because we're playing games on phones instead of PCs or consoles, the HPC community would have to bear the cost themselves. The cost is enormous.
The only reason NVidia engaged with the HPC community in the first place was a reduction in PC sales.
ARM is fine for what it's intended for, but a fire-breathing high performance general purpose CPU suitable for weather forecasts it is not.
Yet. The Mont Blanc Project ("European Approach Towards Energy Efficient High Performance — thank goodness they didn't try to bludgeon that into an acronym) is addressing just this issue, using the Cavium ThunderX2™ mentioned in the article. This press release is a bit more readable than the project's website as a whole.
"This is huge. SBSA is the real threat to Intel."
I will agree at least to that extent. For years, ARM has been a fixed-hardware architecture: hobbled for the most part by those vertically-integrated black boxes. To endorse and encourage the use of a general enumerated bus opens the way for ARM systems to be more general-purpose since ARM CPUs no longer have to, as the article notes, be bound to fixed hardware profiles.
"For years, ARM has been a fixed-hardware architecture: hobbled for the most part by those vertically-integrated black boxes."
Certainly that's part of it. But there's more to this than vertically integrated black boxes.
Who remembers the mess that was WinCE and its derivatives in the PDA (and allegedly in the set top and other embedded) market? PocketPC and HandheldPC2000 and so on.
Some of these designs were not "vertically integrated black boxes", some of these were meant to be able to run an OS and built-in apps plus whatever else the customer might fancy.
How well did they work out for MS and the MS-dependent hardware vendors who bought the Kool-Aid?
Was it ARM (or MIPS or ...) that hampered those things or was it more about Windows CE's all round uselessness, which just served to emphasise that neither Intel nor MS could do without each other.
Weeeellll... let's give it a couple more years, shall we?
I've still got a Jornada 720 somewhere, see e.g. http://www.hpcfactor.com/reviews/hardware/hp/jornada720/
Decent (albeit not Psion-class) hardware for its time, crippled by Windows CE-derived HPC2000. In comparison, just think what Psion could have done at that stage, if they'd had a chance to compete fairly against MS.
Yes, it's been problematic swapping an EXISTING install to to a different vendor's ARM system, in comparison with basic BIOS, then ACPI, then EFI based boot commodity x86-64 motherboards or even laptops.
This also is the route not just for servers but ARM based convertible tablets (with keyboards) or ultrabook style ARM running iOS to replace MAC OS, or Linux to replace Windows. At present it's trivial to download and install Linux from a USB stick to an x86-64 based laptop, but needs rather more planning and customisation to install Linux instead of iOS / Windows / Android / Chrome on an ARM based tablet, or Linux based OpenWRT on a router.
I look forward to the day of being able to customise a Linux distro for TVs instead of the garbage inflicted on users called "Android TV".
Great article.
I look forward to the day of being able to customise a Linux distro for TVs instead of the garbage inflicted on users called "Android TV".
This sounds very much like wishful thinking to me: in the consumer space Android has pretty much beaten Linux off because most consumers don't really relish the idea of customising the software on their TV. They want the easiest access to their favourite shows which will always come with some kind of DRM.
"in the consumer space Android has pretty much beaten Linux off because most consumers don't really relish the idea of customising the software on their TV. "
So how do you explain the fact that consumers really relish the idea of customising their Android (based version of Linux) software on their phones?
So how do you explain the fact that consumers really relish the idea of customising their Android (based version of Linux) software on their phones?
Because they don't 'customise' it, consumers merely do: a little tailoring (change settings), furnish (add app's) and decorate (change the wallpaper) to personalise their device...
'Customise' implies doing more, such as taking a Nook Simple Touch eReader and turning it into an e-Ink Android device - which doesn't mean that yesterday's customizations don't and can't become tomorrow's stock apps and features.
"So how do you explain the fact that consumers really relish the idea of customising their Android (based version of Linux) software on their phones?"
Dragging icons about, adding a couple of widgets and changing the wallpaper isn't really that much of a customisation. And I bet most people don't use widgets.
"Yes, it's been problematic swapping an EXISTING install to to a different vendor's ARM system"
Different vendor's ARM system? Make that [same] vendor's [allegedly slightly] different ARM system. The Raspberry Pi 2B went from v1.1 to 1.2 swapping the CPU but not the rest of the SoC from 32 bit to 64 bit with the consequence that images which will boot on the old 32-bit model 2 and images which will boot on the 64--bit model 3 fail.
As I've spent the weekend discovering.
@AC
You're missing the point about SBSA here. Yes you can retarget to a different hardware platform if there is already a port for the platform . However if your platform is not supported then you are out of luck. However with SBSA you no longer need to target a platform, you just target SBSA and it will work on any target that supports SBSA if there are suitable drivers available.
I completely agree with the importance of SBSA for Linux, but think the dead of Windows as predicted in the article is nonsense. Windows on ARM is a server-class products; the sysadmins and developers working with it will understand that it doesn't run x86 binaries.
MS' real problem is licensing. MS need to find a way of making ARM licensing as fair as x86 licensing without impacting revenue. This is difficult because ARM isn't as capable core-for-core as x86 but it might be capable-enough for many uses. Do they over-charge for ARM or lose revenue by scaling down the cost for ARM? Or do they make a single-user license free and bump up the CALs?
Windows may not be the future, but there is still a stack-load of profit in it. I think MS have miscalculated. If people are going to rewrite apps for the cloud, I think they are more likely to go AWS. What MS should have done is put effort into useful server stuff. Work with the vendors so that they can interrogate the power supplies in the servers and routers so data-centre power management becomes easy. They should have made some decent load-balancing - perhaps worked with Intel to build hardware load-balancing into NICs. They could have shifted their server pricing model to opex, rather than promoting cloud, which will eventually eat their lunch. They should have done "Automation for Small Business" (on-premise) where latency and accommodating legacy applications is key. Their server products focused on the large enterprise at the expense of making things "cloud-easy." A tweak to their licensing model away from per-core and per-cpu and they could have sacrificed performance for ease-of-use and made all those SaaS threats go away.
If SQLServer for Linux actually becomes a thing, it will morph into Postgres for Linux. First for the less critical applications (which will pre-package it) and then for for the more important stuff.
I smell the whiff of burning platforms. It is still quite a long way off, but it is there.
Instead, they were faffing around with Vista, Windows 8 and 10.
It's those I fear most.... the average application quality is far worse than it was before "web native" learnt how to use a keyboard. Most applications are nowadays built upon layers and layers of half-baked frameworks, and the "web native" is always attracted like a moth by the last shining one. More and more "applications" are really ugly, uncomfortable to use, slow, and ill-designed. But of course they contain all the latest buzzwords. IoT is a perfect example of where "web natives" will drive us all.
MS killed itself in the server area with its licensing policies - were CALs really needed after you bought an expensive server license, plus the client OS ones as well? Linux adoption was often driven by purely financial reasons - it was cheaper of free - despite replication many of the Windows features - i.e. AD - requires to cobble together many disparate pieces and pray them work together. Of course, if your business is just to publish thousands of cheap web hosting servers, that was not an issue.
But Linux is often a step backwards - while its kernel is good, a lot of the stuff built upon it is still truly ugly, developed by teams lacking the proper resources, and addressing only their subset of features - the only good thing being usually the commercial applications ported from Unix, which was another reason to replace those expensive Unix licenses with the cheap Linux.
So yes, in the long run the cheapest win. You also get what you pay for...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
