Is MinWin really the new Windows 7 kernel?

Is MinWin really the new Windows 7 kernel?

Summary: Many of the articles and blog posts I've read about Windows 7 in recent days mention MinWin, usually following up with the observation that it's the new lean kernel that's going to be at the heart of Windows 7. I think this conclusion is wrong. If you don't believe me, go back and watch the eight-minute video snippet that got this all started last fall. Too busy to spare eight minutes? No problem - I've transcribed the relevant parts here.

SHARE:

In my post about Windows 7 back in January ("Windows 7 = Vista Release 2"), I ended with a statement that deserves some follow-up:

Those who are predicting that Windows 7 will include some radically stripped-down kernel (the so-called MinWin project) or a new file system are missing the point completely.

Many of the articles and blog posts I've read about Windows 7 in recent days mention MinWin, usually following up with the observation that it's the new lean kernel that's going to be at the heart of Windows 7. That misinterpretation was widely reported in October, when this story first appeared, and it’s become conventional wisdom since then. Australia’s IDM, for example, included this authoritative-sounding statement in a March 17 report:

Windows 7 … is being based [on] a completely stripped down and re-vamped “MinWin” kernel that will be tightly knit with its Windows Live services.

Or this March 5 report from TechNewsWorld in the U.S.:

Windows 7, the next version of the Windows platform, which will be the successor to Windows Vista, has its own kernel, MinWin.

Even my estimable colleague, Mary Jo Foley, thinks MinWin is in:

Microsoft — last we heard — had managed to create “MinWin” — a stripped-down [version] of the Windows kernel, that it is planning to make the heart of Windows 7 and its successors.

I think this conclusion is wrong. About the last thing Microsoft wants to do in Windows 7 is make radical changes at the kernel level. (You think they want the potential for another round of driver and app-compat issues like the ones that plagued Vista for the first six months after its launch?) I’m willing to bet that the Windows 7 kernel will contain only minor changes from the one found in Windows Server 2008 and Vista SP1.

If you don't believe me, go back and watch the eight-minute video snippet that got this all started last fall.

Long Zheng pulled this excerpt from Microsoft Distinguished Engineer Eric Traut's presentation on virtualization technology last fall. Because the blogosphere has such a short attention span, it’s likely that a lot of people writing about this stuff have never actually watched even this excerpt (the full video of Traut’s talk runs over an hour, and this discussion occurs more than 40 minutes in). I’ve never seen a transcript of these remarks posted anywhere, either. So, in the interest of science, I went back and watched the whole thing again, transcribing the good parts. I’ve bold-faced the key statements here.

Don’t skip the intro, which is pure comedy gold. After you get past the demo of Microsoft Bob running on Windows 3.11 (seriously), Traut calls out the lightning fast boot times of Windows 1 and 2. "Paint. Looks just like PhotoShop, huh?" After that bit of of ancient history (and presumably the irony-impaired will notice the humor), this discussion begins at 3:30:

This is the core of Windows 7. This is a collection of components that we've taken out. A lot of people think of Windows as this really large, bloated operating system, and that may be a fair characterization, I have to admit. It is large. It contains a lot of stuff in it. But at its core, the kernel and the components that make up the very core of the operating system actually is pretty streamlined.

He’s talking about the guts of Windows 7 here, but he could just as easily be referring to any other recent-vintage release of the Windows NT family.

It's still bigger than I'd like it to be, but we've taken a shot recently at really stripping out all of the layers above and making sure that we had a clean architectural layer there, and we created what we call MinWin.

Now, this is an internal only - you won't see us productizing this - but you could imagine this being used as the basis for products in the future. This is the Windows 7 source code base, and it's about 25 megs on disk. Compare that to the four gigs on disk that the full Windows Vista takes up. We don't have a graphics subsystem other than text in this particular build, so you can see that's our Windows flag [referring to an ASCII art splash screen].

It’s a research project, not a product plan. They started with the Windows 7 source code base and fiddled with it to see how small they could make it and still have it run something. It might turn into a product someday, but it’s certainly not going to be in Windows 7.

OK, now skip ahead to about 7:10 or so:

[I]t'll be a while before you can build something directly on top of this really tiny core. … Like I said, we don't have any productization plans for it. We're definitely going to be using this internally to build all the products that are based on Windows.

Traut points out that Windows Server 2008 is “based on this thing called Windows Server Core”:

[I]t basically ships in this minimal state - and when I say minimal, it's about a gig and a half on disk, which is a still a lot bigger than what we're talking about with MinWin - but it has all the stuff necessary to run the different server workloads. … I mentioned a gig and a half on disk; that's still pretty big. In memory it's much smaller. That contains all the drivers, many of which you don't need for any given platform.

And there’s the real take-away from this presentation. The immediate goal isn’t to trim Windows 7 to something that will fit on the head of a pin. Just remove some optional components and make the whole thing smaller and more modular than Vista, just as Server 2008 is smaller and more modular than Server 2003. If Vista’s successor in Windows 7 can trim its core space (and memory) requirements by a third or half, giving you the option to choose which components you want to run on top of that relatively lean base, that will be impressive enough for all but the most vocal Vista critics.

Update: For more on this subject, check out this excellent Channel 9 interview with Mark Russinovich. At about 14:40, he addresses the confusion over MinWin and the kernel: "The word 'kernel' is used loosely to mean a whole bunch of different things. ... The NTOS kernel is the core of Windows that runs in kernel mode, and it's got a lot of support components around it that also run in kernel mode. And then there's layers of system level components that run in user mode but are still a part of the core OS.If I've used the word 'kernel' around MinWin, I'm really talking about the core of the system."

Topics: Software, Microsoft, Operating Systems, Windows

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

43 comments
Log in or register to join the discussion
  • I think I'd be ready to bet the farm on that ...

    "I?m willing to bet that the Windows 7 kernel will contain only minor changes from the one found in Windows Server 2008 and Vista SP1."

    I agree with you there Ed. As cool as MinWin sounds in theory, in practice it would hobble Windows 7 in much the same way that Vista was for a good six months due to compatibility issues.

    But ...

    "If Vista?s successor in Windows 7 can trim its core space (and memory) requirements by a third or half ..."

    Now if you were going to offer me a bet on that (and I noticed you didn't), I'd take you up on it. While I can see services and processes that Microsoft could cull from Windows to make is sleeker, I doubt that it could cut the memory requirements by 20 - 25%, maybe more likely 10 - 15%. Half, I really doubt that.
    Adrian Kingsley-Hughes
    • No bet!

      I think you're right. Although I'd be curious to see how small a business-focused Windows Basic version could get.
      Ed Bott
      • Or gaming ...

        ... another market that would appreciate fewer services running by default would be gamers. True, the gains would be small in reality (for all the yak on the web about disabling services and so on, all the testing I've done shows that it makes no difference in most cases as games are mostly throttled by the video card), but it would be a welcome nonetheless.
        Adrian Kingsley-Hughes
    • i'll bet they could, if they tried really hard...

      Just like there are fully functional modern Linux distros that run on one third or less of the resources that the big distros require... if a dedicated team of devs spent the next 18 months hacking at Windows 7, i'll bet they could cut the resources by way more than half... might even git it running on my old pIII-450.
      brokndodge@...
      • ...if they tried really hard...

        I think getting the memory requirements down to about half what they currently are in Vista could happen only if Steve Jobs was in charge over there in Redmond. He can get people to do amazing things, almost miracles. And that's what Windows 7 will need. See this Wired story for what I mean: http://www.wired.com/techbiz/it/magazine/16-04/bz_apple?currentPage=1
        fakejake3
      • Good point

        I keep seeing people posting things like - just buy a new PC - hmm do they work for companies that don't have set up costs, learning curves for end users, and hardware costs? None of them seem to explain why any company should upgrade to Vista for improved client efficiency or return on investment. If I replace 2000 PCs just so I can run Vista - are my end users able to perform twice as much work? Oh wait - I would need to upgrade their PCs just so they can do on Vista what they did on their three year old PC. Hmm maybe there is some new software they can run on Vista that would justify the expense - nope, I can't justify playing Halo2 to the board of directors as a good reason to migrate to Vista.

        I'm hoping Windows 7 is a lean mean fighting machine - that has fast network copies, fast local hard disk copies, fast boot ups, fast shutdowns and fast application loads. Sure it should be secure but not because it is sow slow no one uses it. Security of an OS seems to go hand in hand with its popularity. I doubt OS2 was very secure but since it wasn't very popular - the hacks weren't as common.
        boed
      • You're forgetting what Windows is ...

        ... Look at the uproar over Vista and you see what people want from Windows. An all encompassing platform that works right out of the box with the minimum of driver hassles and additional downloading. It has to plug'n'play every device ever made, even some piece of junk made a decade ago. Everything has to be boiled down so it's simple, and preferably be wizard driven.

        What Windows is, by default, is a monolithic OS. Windows Server 2008 goes some way to change that but I can't see that methodology being applied to the desktop. Remember, the ideal setup process involves having to click on one button, and preferably people would like that clicked before the PC left the factory.
        Adrian Kingsley-Hughes
        • Not quite monolithic.

          Yes, Microsoft sells new versions of Windows (and the hardware to run them) by increasing the functionality and capabilities of the operating system. But the company also sells different versions for different purposes, adding and removing functionality/capabilities.

          Vista Home Basic, for example, is a version which can run on older, slower computers. And the Vista business versions are graded according to what a customer might require.

          If enough buyers require it, Microsoft does respond.
          Anton Philidor
          • MS hasn't responded yet!

            [i]Vista Home Basic, for example, is a version which can run on older, slower computers. And the Vista business versions are graded according to what a customer might require.

            If enough buyers require it, Microsoft does respond.[/i]

            The customers are asking for Vista to run on the same systems that XP runs on. Now I can understand that new technology needs to be supported. However, the vast majority of people would be willing to upgrade to Vista, if it would run on the same hardware. There is absolutely no need for a system so bloated that it has to have a dual core 2ghz chip with more memory than god has money, just to boot up.
            brokndodge@...
          • yea i would have wished xp would have run on my 98 box or 95 run on my

            yea i i would have wished xp would have run on my 98 box or 95 run on my win 3.0 box.
            thats just the way it is. if your system is running with xp and your happy with it wait till you buy a new computer with vista preinstalled. and as far as bloat go's microsoft is giving people what they want. we as IT pros want one thats slimed down but the ave Joe wants all the bells and whistles.
            SO.CAL Guy
    • Bet

      I thought Internet based gambling was illegal?
      1g2j
      • Not on April Fool's Day ...

        nt
        Adrian Kingsley-Hughes
      • Adrian and I will meet in Antigua later today (nt)

        (nt)
        Ed Bott
    • What's the point of trimming the foot print?

      I'm not saying it's a bad thing, but even if Windows 7 used 1/2 the ram of Vista, would it really matter? I bought 8GB of ram a week ago for 140 shipped. Now you could buy it from around as little as $100AR.

      By the time Windows 7 comes out, I suspect that 16GB of ram will cost no more than $100.00...possibly much less.

      No matter what, memory won't be a huge issue.

      What's important is maintainability, and this process will surely make it easier to maintain windows.

      Surely that's more important than saving a few little ram or HD space.
      notsofast
    • Adrian, I think you are right here.

      I doubt Microsoft would be able to make a complete kernal change, for the exact reason you and Ed said. Compatibility would suffer, much like it did in Vista.

      On the memory, I am a little more optimistic. I think that if they say they will cut memory use by a third, they may actualy cut it by a quarter, which mind you would be nothing to turn the nose at.
      gene_fitz@...
  • RE: Don't expect MinWin in Windows 7

    I saw the video from the university talk and I understood that the kernel has no immediate placement inside a product, for example the next Windows. But they didn't deny that it could become something greater.
    setatakahashi@...
  • Praying, hoping and wishing it will be better than Vista

    A couple of things concern me - unless they improve the copy algorithm and or DRM issues in Vista I still think it will be part of the base core.

    I'm also concerned by your comment about Vista having issues with drivers for the FIRST six months - did you forget about the problems with drivers in the second and third six months? While it isn't as bad as it was when Vista was released nearly a year and a half ago - drivers are still an issue with Vista.

    I really want a better OS than XP (which isn't bad - just antiquated) I have some major doubts about Windows 7 being what I hope it will be. Ballmer ignores any comments about Vista being slow and the MS Fanboys just don't want to admit that it is slower than XP. Some say - what do you expect it is newer so it will be slower - so should cars get slower every year or require jet fuel just to do what they did previously. Should cars consume more gas to get from point A to point B every year - NO. They should be MORE EFFICIENT than previous models.

    In theory if they optimize the code for Vista - it should be significantly faster on new hardware than XP. XP wasn't really designed for multicore or tons of memory etc. Vista should have taken advantage of the new hardware and blown XP out of the water- instead it is the other way around. You can say they had many years to optimize XP - maybe so but are all the XP programmers dead - didn't anyone who worked on XP bring anything to the table when they developed Vista?

    No one wanted to admit Vista was a turtle during development so the emperor is running around naked. Since he now refuses to admit he has no clothes; we are going to have to do something pretty drastic to get his attention ??? apparently the class action suit still not enough to get his attention.
    boed
  • Sounds to me like MinWin is already in.

    "This is the Windows 7 source code base" -- note how he says "the" code base. Yes, all of a given Windows version is a single code base (Paul Thurrott has written about the branching mechanics). I'd interpret this to mean that Windows 7 already has and uses MinWin.

    Regarding productization -- I think he's simply saying that no one will be able to get *just* MinWin (like how the Linux kernel is available independently). He's not saying that MinWin won't be present in Windows 7.

    And regarding compatibility -- If all they're doing is refactoring, moving things around, and drawling lines between components, it's quite likely that they can do this without hurting compatibility: as long as the external APIs are the same.

    And about this gem: "???MinWin??? kernel that will be tightly knit with its Windows Live services." -- Ha! A kernel knit with Live? Whoever wrote that should find a new line of work.
    PB_z
  • Windows 7 = Windows Vista SP2

    Windows 7 = Windows Vista SP2
    qmlscycrajg
    • Do you follow the links?

      That was I wrote in January, and I even linked to that title in the first paragraph of this post...
      Ed Bott