Channel 9 has Dave Probert (who used to be a UNIX head but is now an architect for Windows) giving a four part video presentation of the Windows NT kernel. He does a very good job of comparing the Windows kernel to UNIX-style kernels and how they tackle the same problems differently.
Video Tour of the Windows NT Kernel
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Hmmm… Windows NT(3.4, 4, 2k, xp, 2003) has been out for how long now? And people still need to be told how its designed?
How about you show us the code and we will figure it out for ourselves? hehe.
If you’re so interested, then why don’t you just download the code for 2K?
“My Friend” told me it was only partial source, he did a grep -r -i gnu . and the only gnu code was the GPL header in make files, which would not build…
Are you implying 2k was leaked?!?!
mms not registered, but gxine takes the url with http isntead, but no audio. does it work for anyone else? xine works fine for other stuff.
yea it was leaked. i even got a copy of it sometime back but it was eating up too much dust on my hardrive
No it was not. NT was leaked not 2k. Yes it was FAR from full source. Yes things like this hurts OSS much more than M$.
If you ask me they probably released it themself, just so they could make ppl switch over to a more “secure” win-version (If you are smart you will not dl the source and look at it, it could hurt you in so many ways…. and I’m not talking about M$ kicking down your door.)
Win2k SPx (don’t remember which SP it was) was leaked; everyone grepped the source hoping to find GNU code…alas, none was found.
There’s a book called Showstopper that chronicles the development of the NT headed up by Dave Cutler (who worked on VMS) and also google for “A Tale of two kernels” that compares linux and the NT kernels and also shows where Linux has copied some features from NT.
I just wanted to say that this video was very interesting for me to see. IMHO He has several points in how computers should work, IE perhaps not designed. Computes should work flawlessly out of the box with the ability to tweak it for personal power-usage.
Besides the video and clear message, I like especially the irony of seeing some UNIX-geek stuff on his desk, as he was a former UNIX man.
Actually watched all 4 videos.
At first, I was very excited about how he began outlining some hidden facts about NT and its VMS heritage (and comparing it to early Unux), but then just as he would come up to something interesting, he’d change topics. Of into the new tangent he’d go to, and again come close to revealing something new that I’d never heard/read about before, and then change topics again. Throughout the entire 4 video clips, he never strays from Operating Systems 101 which is presented in all computer science courses. He constantly switches focus before saying anything new. Disappointing.
one of the interesting things he does present is that Microsoft are moving away from the global kernel lock into much finer grains (along the similar paths taken by FreeBSD, Tiger and the Linux 2.6 series).
/me hugs his BeOS box.
mr. morot wrote:
“No it was not. NT was leaked not 2k. Yes it was FAR from full source. Yes things like this hurts OSS much more than M$.”
2000 is based in large part on NT code. Ever read the splash screen when a Windows 2000 server boots up?
“Built on NT technology.”
Yes exactly.. BUILT on, not IS.
Anyway, I was wrong. It was NT4.0 and 2k that was leaked, though only partial.
My bad, I stand corrected
Copy/Paste isn’t all it’s cracked up to be.
This is probably a pointless debate. If you can see the actual source code or read about the origin of a project, you can do an elaborate ‘geneology. See http://www.levenez.com/unix/ for an exhaustive history of the *nix family.
I’ll bet that ‘build on’ means that at least 90% of the NT code base is in Win2k. But, the Win2k probably doubles the code base of NT, so there is plenty of new code in each version of Windows. Presumably the work on the kernel is largely ‘polish and refinement’: bug fixes and finding algorithms that scale better (much like we have been seeing in Linux 2.6, NetBSD 2.0, FreeBSD 5.x and OpenBSD 3.5+).
Hopefully, MS is also going through the exhustive code review process that made the OpenBSD team famous. (Creating a new protocol is fun, implmenting it cleaning is the hard work.) But, since we cannot actually see the source code from each version of Windows, we will just have to trust that Microsoft is doing all of these things. Verfiying claims about closed source is a difficult task.
These are really really low quality videos…
Come on people, Windows 2000 was Windows NT 5.0 until marketing got their hands on it. You all know that straight version numbers are unpopular with closed source now. MX, CS, XP and so on. Get a grip, it is NT.
He talks about NT vs Unix on a kernel level, and to show that Unix is out of date he complains about vi? I’m sure that NT is a beautiful kernel, I’ve never heard a bad thing about VMS. But, this guy seems like a typical Microsoft spokesperson:
1.) Good personality.
2.) Never runs out of jargon.
3.) Doesn’t really make flowing arguments because they can be argued with.
Maybe it improves after the first video?
no wonder windows is such a mess. its kernel architect cannot express his thoughts clearly, consisely and to the point…
The guy hits the nail on the head…sending email was that crazy? Thank god for Windows…now dont get me wrong I know it has drawbacks and security issues but you cannot deny Windows getting way better than ever. M$ has the unique ability of making things so easy for users to use.
I’ve read a few times that they’ve rdone the whole NT kernel for longhorn and it’s brand new, I don’t know how true this is or not but it would be nice to see performence wise the changes made.
The NT kernel as is can hang with the best of them, the only thing killing Windows is the bits on top that fuck with everything, like IE for one.
I hope the new model they are working on for the future works better.
I hope so Anon cause that is what I have been hearing. And if they are going over the NT kernel right now…probably that is what is going to be in Longhorn…a revamped version I hope. I hope all the .NET managed code is worth all the hype cause Longhorn is going to be built entirely from that!
Microsoft definitely went over the kernel, but that doesn’t make it a new kernel. Hell, if they didn’t do a good house cleaning I would be more concerned. Now think about it, the kernel is big, has lots of functionality and bugs that you can’t remove without braking backwards compatibility. Why would you write all of that from scratch if you don’t have to? If you want to know how hard it is to write an NT compatible kernel ask the ReactOS guys.
So that means same old weaknesses and instabilities and kludge and cruft in the Longhorn as well as long as you dont install some 3rd party tools? Looks like Longhorn will be good for one thing only…gaming.
So what is wrong with refactoring existing code? I am sure they go through a process of modeling based on existing and desired functionality… then save what they can, rewrite what they must…
No, Windows 9X was a kludge, NT isn’t. The architecture is good, and there is lots you can do to improve. Not every bug is a pre requirement for backwards compatibility, some backwards breakage is acceptable, and I’m sure there is dead code to remove and so on.
Personally I never was that hyped about Longhorn. All this new stuff smacked on top with .Net seamed a bad idea. I’m sort of glad that lots of what Longhorn was supposed to offer died on the vine or got pushed of for the next version.
You NEVER throw away good code unless you plan on going out of business, or greatly lengthening the time required for development.
I do not think to have a robust kernel you need gigs of space nor do you need craploads of lines of code. I read somewhere that the Linux kernel as an example is just over 4 million lines of code. I have heard M$ Office has over 10 million lines and some similar number for XP?! Its all spaghetti code. Why dont they do a refresh of the kernel and start from scratch? I am sure M$ has a crapload of PhDs on their payroll who can do this. Sure they can continue developing Avalon, Indigo and .NET on the side but man they need to clean stuff up before slapping more kludge on. Me thinks in the end XP will be more cleaner than Longhorn and thats sad if that comes out to be true!!!
The kernel is only a small part of the operating system and the least viable. The size of modern windows has little to do with the kernel. I don’t understand where you got the idea that the kernel is spaghetti code but you are wrong; and as to re-architecting it, I doubt that MS has lots of people better at Operating System design then Dave Cutler.
As to the future of Windows, Longhorn and so on, lets wait and see.
I read somewhere that the Linux kernel as an example is just over 4 million lines of code. I have heard M$ Office has over 10 million lines and some similar number for XP?!
You’re comparing the linux kernel and Office? HAHAHA. Then “I read somewhere…”
Get a clue, you’re making bad assumptions based on bad information.
I think this guy is marketing not kernel architect. I’m pretty sure it’s not Dave what’s his name who helped write VMS.
This is however, the reason why Microsoft has had problems with it’s customes: It doesn’t really care about the service it provides them, it just tries to make sure to attract the developers who write the 3rd party apps for those customers
. (<- My generic opinion of Microsoft is that they exist for developers and not users).
Anyway, I kind of like the videos, they are fairly interesting. But he seems to be of the understanding that linux works the same, under the hood, as System 1 did.
He lost all credibility when he said this:
Microsoft has brought the price of PC’s down. It has not! Stop taking credit for the hard work of Computer Engineers in all those part companies who’ve brought down the price of the computer while you have increased the price of your software!
*sigh*
Microsoft marketing has a nack for angering me.
Microsoft has NOT really even increased the price of Windows since Win 3.11.
What his point was is that Windows is very portable, and that they were ready to have versions of other processors and such, thereby keeping the price of the hardware down, as all of the hardware will be running the same software, makeing competition in the hardware realm.
Basically, his point was, that since Windows has always been the most compatable (hardware wise) OS, it has kept the cost of hardware down quite a bit.
Now, I’m not saying it is true or false, I am just telling you what he was saying.
…act by midnight, and keep the virus scanner as our gift!
Yes, this is a great marketing piece. He compares *NIX of 15 years ago to NT/Win32 of today/tomorrow?!
Multi-threading and SMP on Windows? Maybe someday, but I’ll keep my *NIX running on 64+ CPUs thank you.
POSIX is subplanted by Win32? (Read: “You will assimilate!”)
Microsoft driving down the cost of computing? Last I checked, Windows 3x cost about $100 on a $2000 computer….now it costs $150+ for a <$1000 computer how does that math work?
I’ve got to go now, time to install MS Office that I just bought for $429.
GPF
Since it hasn’t been said yet on this thread and since some did not watched the whole thing because of that, let me warn:
If you already have notions about how an OS works, don’t waste your time and skip the first three parts. They are not NT specific, only the usual introduction for memory mappings, scheduler etc… I remember back in school where we had courses on openvms, unix and win32 we had 3 times the same introduction (the DOS one was quite different ;-). Same thing here, but this is quite a good reminder though 😉
The 4th part is about the micro kernel design, extending APIs using services etc but nothing really new though.
You know, all these things we will be able to do with the HURD … … … … LOL! 😀