Linked by Thom Holwerda on Thu 23rd Dec 2010 19:19 UTC
Windows The rumours about Windows possibly being ported to ARM has left a lot of people bewildered; why would you port Windows NT when Windows CE 6.0 is a perfectly capable operating system? Putting all the pieces together, it's actually quite clear why you would want Windows NT on ARM: servers.
Thread beginning with comment 454621
To read all comments associated with this story, please click here.
Bill Shooter of Bul
Member since:
2006-07-14

It doesn't make any sense. Windows rack mounted servers ( the kind that would benefit from reduced energy consumption and heat output) are typically used by larger enterprises. Larger enterprises that *have* written software for windows server, and would have to port it over to ARM Windows.

PLus, I've heard the ARM server rumours as well, but I don't believe they will be a viable option. There is today, not a market for ARM based servers. If you are moving from the mobile space, it makes a heck of a lot more sense to gradually move up in processing power: Ie desktops/net tops. Even tablets make more sense: they exist already and use ARM predominately. Microsoft has been abundantly ( and stupidly) suggesting over and over that their tablets will be full windows 7, not CE, not zune OS, not windows phone 7 OS.

Basically, I think anyone suggesting CE on tablets is just giving Microsoft too much credit to make a good decision.

Reply Score: 2

hugh Member since:
2010-10-20

"Larger enterprises that *have* written software for windows server, and would have to port it over to ARM Windows. "

A lot of that is probably done in .Net by now...

Reply Parent Score: 2

vivainio Member since:
2008-12-26

Larger enterprises that *have* written software for windows server, and would have to port it over to ARM Windows.

In addition to the .net ancle already mentioned, "porting" means just recompiling. That may not be trivial for ad hoc hackjobs abundant in microsoft shops, but big hitters (oracle, ms itself, adobe, ibm...) will have their software up and running in no time (most of the time going for verification).

Reply Parent Score: 4

Bill Shooter of Bul Member since:
2006-07-14

I wouldn't be so sure about adobe, they can be a bit slow on the use of new platforms.

Reply Parent Score: 3

toast88 Member since:
2009-09-23

In addition to the .net ancle already mentioned, "porting" means just recompiling. That may not be trivial for ad hoc hackjobs abundant in microsoft shops, but big hitters (oracle, ms itself, adobe, ibm...) will have their software up and running in no time (most of the time going for verification).


[ ] You have already done that before and know what a tedious and difficult job it can be.

Let me remind you, that Adobe and Sun took years to properly port their plugins to AMD64 which is "just" another platform in your humble opinion. Way more software nowadays still uses handoptimized code and leverages features which are only present on certain architectures making the code very difficult to port. Ask yourself why there is no native version of Skype for AMD64 yet and why the JIT compiler of Chrome and other browsers were available on x86 first.

Porting code to another architecture by just recompiling the code only works if you wrote your code with portability in mind right from the beginning. Unfortunately that's not the case for many popular closed source software applications, especially when the codebase is older than 10 years.

Naturally, FOSS projects usually are way more portable since they were already designed with portability in mind.

And no, Microsoft certainly can't just easily recompile the whole NT codebase, which is already 20 years old, easily for ARM. Hell, ARM is a vast different architecture, it differs much more from x86 than AMD64 even though the latter also isn't just an x86 with doubled data and bus sizes (which isn't even true since all current AMD64-compatible CPUs feature 48bit addressing only).

ARM is for example big endian as compared to Intel's little endian and it's a RISC architecture as compared to the CISC one of x86 CPUs.

Adrian

Reply Parent Score: 3

Vanders Member since:
2005-07-06

PLus, I've heard the ARM server rumours as well, but I don't believe they will be a viable option. There is today, not a market for ARM based servers.


Yes there is. It's not going to replace x86-64 overnight, but Linux deployments make up a significant portion of server sales. Linux deployments mostly don't need to worry about the CPU architecture.

I know I could replace a significant number of x86-64 servers with ARM without any problems, and the power savings are likely enough to make it worthwhile. I don't believe my situation is at all unique.

Reply Parent Score: 4

Bill Shooter of Bul Member since:
2006-07-14

What do you mean by "Without any problems"? You haven't benchmarked any of the ARM servers for your load, because they don't exist. How do you know these mythical servers will be able to handle the current load of your x86-64 servers?

Reply Parent Score: 2

galvanash Member since:
2006-01-25

Counter-point... One of the largest growth sectors in the server hardware market is large data centers for cloud computing. What you want in such data centers is small, cheap, low power (critical for operational costs) servers that generally do one of two things:

1. Run custom written distributed database/file storage software.
2. Run custom written web servers/caching proxies that are purpose built for cloud applications.

Linux practically owns this market as it is now, and ARM servers are more than likely going to be very attractive assuming they offer advantages for power use/thermals/size/etc. Since Linux is already the defacto standard, ARM has no real obstacles in its way to enter that market, so it is likely to at least get its foot in the door. Microsoft would like a piece of it. Simple really, I frankly don't see how they could ignore it.

Porting Windows (in one form or another, I'm not claiming it will be exactly the same as on x86, but CE is simply too striped down for this type of use) to ARM would in effect give them somewhat equal footing once the hardware becomes attractive enough that customers start buying it. True there is something of a challenge in porting the software required for these types of deployments, but it isn't inconceivable that Microsoft could offer some advantages here that might sway some customers (i.e. existing Microsoft shops wanting to build purpose built cloud services).

Point is without any presence on ARM - if ARM servers take off Microsoft is completely locked out of a potentially huge growth market. Do you really think they are willing to let that happen?

Reply Parent Score: 5

lemur2 Member since:
2007-02-17

Point is without any presence on ARM - if ARM servers take off Microsoft is completely locked out of a potentially huge growth market. Do you really think they are willing to let that happen?


This was my take on it as well. New power-efficient ARM chips that could be used in a desktop or server role are on the horizon, and they will be available to market in a couple of years. Some OEMs are sure to take advantage of such chips and make a server which achieves a significantly better power-per-performance metric than x86 or x86_64 can, perhaps it will be more than an order of magnitude better.

Currently, the only viable server OS for such a machine would be Linux.

IMO, Microsoft cannot tolerate that. If Microsoft want to stay in the server business in a couple of years time when these machines come to market, Microsoft will HAVE to have a Windows server on ARM product ready to roll. If they are even a little tardy, Linux will take over this entire market segment.

Reply Parent Score: 3

fasted Member since:
2006-11-09

If you read this article http://www.linuxfordevices.com/c/a/News/ARM-CortexA15-MPCore/ , it explains the larger reason's for the future demand for this. Energy savings, and virtualization.
First, even the appearance of being a "green" company could bring large business your way. Think of Google's effort's over the last few years to cut down on their carbon footprint. It's saving money ,also, as well as the environment . Then think of the virtualization solutions on big iron servers, where the base operating system ( Linux, most likely) host's any other operating system that is required . Take a look at the top 500 fastest super computers to see how many aren't Linux, and you'll see why this could be very important to Windows. IBM was king of the big iron at one time, a mistake Microsoft doesn't want to see happen to it either. Also, imagine the sync capabilities of your tablet, phone, home and server being all one type of chip and Operating system. Much more at stake than just iPad's and iPhone's. Just my two cents.

Reply Parent Score: 5

Nagilum Member since:
2009-07-01

I do not agree, as an Admin of such a company I see the majority of these servers run IIS for asp/aspx, MS SQLServer or Java.
All that stuff will easily run on ARM.
Many Webservers out there only act as proxies providing SSO for WebLogic and other backends.
These webservers would run just fine on a dual core ARM with 2GB RAM.
For DB an application and database servers you still will want maximum CPU and memory.

Reply Parent Score: 1