Wednesday, September 24, 1997, San Diego, CA

MR. Allchin: Ever gone to a high school or college reunion? When you go there, you remember the most popular people and the prettiest.But when you get there, you find that often what you remembered isn’t exactly the same now.And often the most popular and the prettiest, they look like they’ve been run over by a truck.And, instead, you find that the geeks, the people who didn’t quite fit in, they’re the ones that continued to develop and grow.They’re the ones who are the most successful and, consequently, they’re the ones who look the best.In the operating system market, it’s easy to get distracted by short-term popularity, by a fad, sort of like high school.

Today, I’m going to talk to you about our operating system strategy, primarily focused on Windows NT, a system that hasn’t won any popularity contests.In fact, Windows NT was developed in a different way from most PC software.An extremely experienced engineering team focused on automation, in terms of testing for code coverage, and we even marketed it the product differently.

I’m super proud to be able to represent the operating system team and give you a tour of Windows NT 5.Now, Windows NT 5 is very strategic, because of all the advances over Windows NT 4.It’s a massive release, the most comprehensive release we’ve ever tried to do.The second reason Windows NT 5 is so strategic is that Microsoft is going to bet the company on it.So, let me start by giving you a little context.

You heard Cam, Chris, and Paul talk about this Windows DNA application architecture.Well, if you notice behind it, it says, distributed operating environment.That distributed operating environment can be broken down into one word, Windows.Now, what I’ve shown on this slide is an architectural view of what NT looks like.Now, there are probably different ways that you could break apart the architecture, but the reason why this particular architecture was chosen is because it matches the way the NT organization is constructed.It’s the way we’ve been able to scale this massive product while avoiding the mythical mammoth.

Now, it’s not composed of just one system.It’s obviously composed of a whole set of systems that work together in unison across the network.These systems working together form this distributed substrate that these three-tier applications can run on.The most important part here is that we’ve always had a belief that the client and server, if we could make them essentially identical, that symmetry would be great for development, because you could test it and we could gain further synergy in terms of our own work, and customers could have a better environment for reducing cost of ownership.So that’s something that we’ve been working on through the years.

Now, there are other people who understand this distributed substrate, and they’ve been working on it as well.There are two approaches you can use.The first approach is to build a little mini operating system that runs on another operating system, and you port that onto all the different environments.And there’s obviously a very good reason for doing that.You have portability.There are also disadvantages.It’s much more complex.You will end up with too many layers of resource management within the different operating systems.So, it’s more complex for administrators and the like.

It’s also an unbelievably inefficient approach, because you are creating a mini-operating system.You guarantee slower performance because there’s just more code that you end up having to process.And, if the operating system is evolving very quickly underneath, you end up basically starving the applications above from taking advantage of the operating system technology beneath the surface.

For these reasons, we’re obviously building as much technology as we can into the base operating system.There’s no reason memory management should have been separate, or networking and the like, and we don’t think transaction processing and the like should be separate either.The operating system is just there to create this distributed substrate so you can run applications on it.

Now, we do put certain amounts of software on other environments.DCOM is a classic example of that.We truly believe that what corporations want is interoperability.So we’ve had a motto for some time of taking any client to any server, any server to any client.We try to use standard protocols whenever possible and, as necessary, we will enhance protocols, create new ones, but we will publish them.We also believe in rich interoperability.So, for example, instead of just worrying about TCP, we try to worry about how calendaring works between these systems.So, our approach is integration and interoperability.

In terms of the Windows family, you’ve heard a lot about Windows 98, and hopefully you all have it and have been testing with it.We see it as an exciting upgrade for consumers.We certainly focused a lot on quality.We think it is a great vehicle for consumers to integrate the PC, TV, and games together.NT Workstation 5 we think is going to be the business desktop.These will be a super-set of Windows 98.We’ll have a good Win 3.1 as well as Windows 95 and Windows 98 upgrades.

There will also be this platform for writing these distributed apps that have been discussed.And, yes, we have two operating systems today, but the one thing you may not realize is, we have greater synergy on the API front between Windows 98 and Windows NT 5 than we have in the past.So, even down at the driver level, we’ve tried to focus and get the samemodel on both of those systems.

In terms of deployment, say in a business, we think everybody should move to Win 32.Why?Because it’s the applications that you’re writing, they’re the ones we want people in corporations, or even at home, to go to.We think you should build Windows DNA apps, and we think corporations should buy them.In businesses, we think Windows NT is the obvious standard, and consumers should upgrade to Windows 98.We do understand that there will be mixed deployments within organizations, and that’s fine, because, again, we have more synergy between the APIs than ever before.

Since you’ve seen Windows 98 and probably played with it, I’m going to spend most of my time talking about Windows NT.Now, we’re never going to get done with Windows NT.It’s always going to be a work in progress, almost a work in love — a work that we love.

So the first thing we did was to try to produce a solid foundation.We did that in ’93 with the microkernel.We focused on size and performance, because if you’ve ever touched Windows NT 3.51, you’ll know that it was slightly large, and slightly slow.We added interoperability in the next year.We then changed the interface and added Web integration, last year in Windows NT 4.0.This year we’ve added enterprise features in the Windows NT Server Enterprise Edition. So the real question is, what’s our focus for Windows NT 5.0, and why is it so important.

Well, the answer is pretty easy.The first is it’s a superset of Windows ’98.That means it has the opportunity to become the volume platform.It’s going to make our lives better, in corporations and for you.We are hardcore onmanageability, in terms of Windows NT 5.0.We think we can drop the total cost of ownership by up to 50 percent with Windows NT 5.0.

It’s also this foundation, for these distributed applications that we’ve been talking about.I’ve been doing networking for some time.I think Windows NT 5.0 will be the most advanced distributed system ever built, that’s been shipping in a commercial space.There’s been some great things in research, but nothing of this magnitude.It’s going to enable a new generation of applications.

Now, some facts.A lot of times when I meet with people they say, hey, I don’t know, are you guys ever going to get Windows NT built, I mean, it’s getting bigger.Well, big isn’t the word for it.And this slide is slightly messed up, so I will explain it to you.Windows NT 3.1 started out as 6 million lines of code.And the slide is moving automatically.It started off with 3.1 million lines of code.By the time we get to Windows NT 4.0 we are 16.5 million lines of code.Windows NT 5.0, beta one, is now 27 million lines of code.Think about it.How are we doing it?

Well, one of the things that you probably can’t see, because this slide is messed up, is that we have about an equal number of testers to developers.We have about 400 testers and 400 developers.And the people who are in tests aren’t running the tests.They are from some of the best schools in the nation.They’re writing tests to actually get code coverage on NT.It’s the way we’ve been able to scale.Not only do we partition the product, but we also align the testers with those components and they write tests.And we have very good ideas about the code coverage that we have on the system.

Another fact on this frame is that, in the X86, if we start from scratch, not doing the incremental builds, which we do most days, it takes 11 hours to complete.If we do an alpha build it takes us eight hours.Just general facts that people ask about.

The other thing that I think is interesting is that the number of outside companies that have offices within the building I’m in, or building 27 or building 10, or any of the other buildings that are on the campus, shows that it really, truly is the number one industry initiative.You know, I walk down the hall and I see somebody from Cisco, I walk down the hall and see somebody from Digital, from Intel, and the list goes on and on and on.So I think across the board, it’s going to have a major impact on your life.

Another thing that developers ask about is, well, how exactly do you go about the process, how do you do source control, how do you control the build.Well, one of the things we do is we do a cycle, where we produce a new system every day, and automatically distribute it.And while we can’t go into a lot of depth on the source control and the like, I thought that we’d show you a video that gives a day in the life of NT.So can we roll that video.

Video voice: The daily cycle for Windows NT begins in the evening, when the newest build is released at 6 p.m.The test team initiates the stress process of thousands of clients and servers, that simulate the work of hundreds of thousands of users.As evening falls, our testing teams settle in.They focus on writing automated test suites.With this magnitude of a product, automation is the only way to achieve the level of testing required.The stress tests are the front line proving grounds for the overall reliability of NT 5.0.

At 5:00 am, the remote diagnostics process begins.The logs for the overnight stress test are carefully analyzed.They are then posted on a Web page, so that any team member can monitor quantitative results, at any time.By 9:00 am, the stress and bug reports are sent to team members.At 9:30 am, team leads, representing every system component, assemble to prioritize which bugs get fixed that day.Over the next several hours, developers begin making fixes.Before they check in any code, they must run functional tests, written by the test team, and submit their changes to dev leads for approval.

In the meantime, our hardware test labs are busy round the clock, sorting out compatibility issues.The lab maintains systems from more than 300 different manufacturers.In addition, each developer builds the next day’s version of Windows NT, using the most recent test release.For beta one alone, we tested 405 video adapters, more than 1,400 printers, 374 net cards, and more than 2,600 modems.Every day we add more devices to the test matrix, so that by the time we ship NT 5.0, we’ll have maximum compatibility coverage.

At 3 p.m., approved code changes come back to the lab, and the team goes to work assembling the new build.Before it’s released the build is installed and tested for integrity.Then it’s automatically distributed over the network.And at 6 p.m., as the day’s build goes out the door, the process starts all over again.

MR. ALLCHIN: We do this every day.It’s the only way to build software of this magnitude, and it works.What I’m going to do, since you’re probably very familiar with Windows NT 4.0, is just talk about the new things, giving you a quick tour of Windows NT 5.0.There’s plenty of drill down sessions, and for everything that I’m going to talk about — any demos I show you, you’re going to see the code in the coming days.

First, in base services, what we try to focus on was basically customer satisfaction.We wanted to drive the system down and solve our basic problems in hardware compatibility, and go up as well, in the enterprise area.Certainly, one of the biggest issues that has been raised to us is, hey, you don’t have plug and play, and you don’t have power management.Well, we will now.We didn’t have a common driver model between Windows 98 and Windows NT.And we will in Windows NT 5.0, in fact, we’ll go a little bit farther than just using the base NT driver model, which is what WDM is, we actually added the ability for signed drivers, and we added instrumentation capability to the drivers.

Of course, device coverage has been a real thorn in our side.And that’s something that we have a tiger team focused on.And working with the industry we will have great coverage at Windows NT 5.0, including all the things that you’ll find in Windows 98, like DVD, and fixing some issues that we had like the absence of scanners.

In terms of scalability, we’ll have 64 bit VLM. For the enterprise version, we’ll also have I20 support, for basically off loading the IO capabilities, so you can actually have a channel there, if you will, and continued focus on SMP scaling.In terms of storage management, we’ll be working on the base file system, with NTFS 5.0.We’ll also be adding FAT 32, directly compatible with Windows 98, or even Windows 95, OSR 2.We’ll have disk quotas, attributed file support, this was something we promised a long time ago, so you can put properties on any file in the system.And then coupled with the fact that we can do real time property and content indexing, you’ll be able to have those attributes available to any user who wants to quickly search their storage for them.

We did more than just work on the file system, though, and the storage area.We’ve added a whole new media management system.And it’s actually multidimensional.We added a multilevel storage capability, so that we will ship an HSM, two- level, and we expect others to add three and four level support.We added shared access through a common API, to get to robotic changers, so in your application you no longer have to be device sensitive and write specific code.You can just talk to this interface, and you’ll be able to talk to any of these devices.

We also added a new back up, with disaster recovery.So once you have this optical, or tape system, you’ll be able to plug it into the machine, and the operating system will put back all the facilities, that the way you tweaked it, plus all your data will go back, all from one back up medium.We worked on hot spot and volume management, so you could dynamically rearrange what’s going on, once you determine where the hot area is on a disk — all very high-end features.

At Microsoft more than 50 percent of our revenues come from outside North America.And we have had a program underway for years to try to create an efficient way to be able to produce systems for this market.And the right way from the operating system side, and from the applications side is to do one single binary for the worldwide market.It takes a lot of capability in the system in order to do that.What that means for you, is that it’s going to simplify your engineering, and we can help move you into worldwide markets easier.With Windows NT 5.0, there is a multi- user — multilingual user interface.

So on one machine you’ll be able to switch the languages that you’re using, because there are language packs that can be added to the system.And you can add a bunch of them on there.It means the same code runs everywhere.That’s true for the operating system, but it’s also true for you.So you want to enter the Thai market?You’ll be able to do it pretty easily with simultaneous support of most languages.

We’ve been asked by corporations for one operating system, and have somebody who speaks German, and somebody who works in French.They want to be able to switch between the languages.They actually have different applications localized in the different languages, and they want them to be able to run on one machine.We did this by having one single worldwide API across all the systems.The best way to show this is through a demo.And Keith White is here, from the NT group, and is going to give us a demo of that.

MR. WHITE: Fortunately, I don’t need any hardware with that, so we should be okay.Basically, what I’m running here is Windows NT 5 Workstation, beta one, and what I’ve done initially is, I’ve logged in using the German character set.So, I’ve logged in as a German user.You’ll notice that the start menu is all in German.If I go to the control panel, everything is set up using the German character set.If I go to an application, such as Notepad.You’ll notice, too, here that all the menus are set up using this German character set.

MR. ALLCHIN: So, that’s great.But that changes the app a little bit, how about input?What can you do about input?

MR. WHITE: Right.See, you’ll notice that even though I’m in German, I can type in, in English, but the other key benefit here is that I can also switch to different types of languages.So, for example, I can type in Hebrew, if I wanted to, and I can even switch to, how about, let’s try Arabic, since that goes kind left to right.So, again, you have the full functioning of the different character sets inside whatever you log in as. MR. ALLCHIN:So the application didn’t have to change at all.Suppose that you want to switch to a different user.

MR. WHITE: Okay.So let me go ahead and log off here.And I’m going to log in as a Japanese user this time.So, again, I’m using the same machine.We’ll go to the log-in screen, and instead of German, I’m going to type in J.And this time, when Windows NT comes up, we should see the Japanese start menu down below.We’ll let the active desktop kick in.And you notice that even the Internet Explorer here is in Japanese.

Now, if we go to the start menu, you’ll also notice that all of the menus are in Japanese as well.So, again, all in the same machine with the same binary.

MR. ALLCHIN: Great.Great. What this means is that Windows NT 5 will be the most advanced multilingual system ever built, and I think what it means for you specifically is that you’re going to be able to take one NT machine and do all the testing you need for all your languages, and then you’ll be able to ship one application sort of independent of what the operating system is there, and let the operating system handle the input method handling as well as the character display.I think it’s tremendously powerful.So, it’s obvious that one of the things that you should take away from this conference is, you should be writing Unicode and following the design rules.

In terms of network services, our goal here is pretty easy.We’re just trying to build it in.We’re just trying to integrate it.Memory management was a mess when it was not integrated.Networking has been a mess when it hasn’t been integrated.So, we’re just trying to wire it in.In terms of the basic support — well, maybe that’s the problem, networking isn’t so basic.We have to focus on simplicity.We’re going to make TCP the default.If it can’t find a DHCP server, it will automatically set that address.And we’re changing the UI.We haven’t pulled it all off in the beta one version, but we will by the time we’re done.

We’ve also unified all the tools with the Microsoft management console.So, instead of having individual tools that you have to deal with, you just will end up with one nice, integrated tool.The other thing is, we’ve made all the networking stacks plug and play.I mean nothing ticks me off more than having to reboot, nothing.

(Applause.)

MR. ALLCHIN: So, from a driver perspective, the system is plug and play, but also from the transport area.Further, the system will have the capability so that it can wake-up automatically when a network packet comes in, whether it be on a LAN or through an ASYNC line, so the system can handle things like faxes and the like.

In terms of the programmable network infrastructure capability, we’ve had some routing technology out for some time.We’re extending that in Windows NT 5 with the capability for multi- casting and unicasting.We also are adding network address translation as a fundamental part of this system.

In terms of RAS there’s a number of protocols.The extensible authentication protocol, the bandwidth allocation, and the like.Now, I’ll just pick on the bandwidth allocation one.Today, when you dial in to an NT machine, you can’t really tell whether somebody is allowed to use multi-link.I mean, when I’m at home dialing in through ISDN, boy, I want 128K Bits per second.Well, everybody else does, too.So the question is, is there a way for the system to decide, per user, whether or not this user should be able to use multi-link or any of the other capabilities of RAS or not.So, integration in with the directory system will be able to determine whether or not this person can use that capability.

In terms of virtual private networks, we’re continuing our work with ATF, Cisco, and the rest of the industry by adding L2TP into the operating system.Another standard that goes along with our PPTP initiative.

Obviously, TCP is fundamental to what we’re about.So, we have a team that’s focused on performance, and adding the newest protocol enhancements in that area, security, so IPSAC and ICMP and quality of service.

In terms of new media, ATM and IRDA and TAPI 3.0 were focused on being able to — eventually, we want to see one wire that can handle video, voice or data, and we want a set of APIs that will give us that capability.In this particular area, adding the streaming capability directly in the kernel will give a lot more capability to the telephony area.

Popping up the next level, the focus of Dave Thompson’s team has been for years to achieve logical centralization, even though the system is physically decentralized.And there are a lot of services that you have to use in order to try to pull that together.Directory services, distributed security, and a variety of other services.

We’ve talked about active directory before.Last year, we gave out an early preview that ran on Windows NT 4, and so you probably got a general idea what it was about.Now, most people when they talk about directory systems, they focus on administration, and making their job easier for this logical centralization.But the other dimensions are just as important.In fact, I think they’re more important.

So, what has a directory server or service done for an end user lately?Not much.And is it easy to be able to extend the system for developers?Those are the areas that we’ve spent a great deal of time on.

One of the reasons why we’re so unique is that we’ve tried to take the best of DNS and X.500 and combine them together.There isn’t a separate database for the DNS table.It’s directly in this — in the active directory.It’s dynamic and secure.It has the capability to extend the schema very, very easily, and many of the hard problems of changing schema in multiple different machines and having them merge and do the right thing are the focus of what we’ve been after for some time. Another thing that I think is really important is that we have tried to focus on scale, and making sure that we’re available.We’re also protocol independent.We really don’t care.You can come in DNS, you can come in LDAP, you can come in MAPI, it doesn’t make any difference how you come into the system.

And, lastly, that point for developers that it’s easily programmable.This is a very high level architectural view of Active Directory.And there’s a couple of key points that I want to make on this slide.There’s so much talk about LDAP.LDAP is the protocol.From my perspective, either good or bad.It’s good because it’s a standard, from that perspective it’s good.But it really doesn’t make up what a directory system is all about.It’s implementation.LDAP isn’t a panacea, it’s just good for interoperating.

Some of the things that we’ve spent a great deal of time on, and frankly we’re not done yet, is design for operations.You know, can the system be backed up online or do you have to take the system down?Can you guarantee that if you crash at any time that you can get back to a stable state and not lose any data?In other words, does it have a transactioning system underneath it.How is the replication?You have a server in Brazil, you have a server here, how is the replication?Does it work efficiently, or can it — won’t it work unless you’ve got a T1 line between the sites?

Another thing that’s absolutely fundamental is that everything is going to change, and every time you sit down and you lay out the tree structure for this directory system, I guarantee you tomorrow it will change.So, we focused on being able to do bottom-up deployment for companies, so that we don’t have to make one massive decision.

And so, just technical problems like doing simultaneous renames across the net, about what happens.You had Joe in the directory system, you deleted Joe, and then you put another Joe back, maybe exactly the same name.Is the system confused?Did somebody get access to some data in the system because it’s got the same name in some access control list somewhere or not? So, what we’ve built is focused on implementation.Underneath the covers, it’s a typed, extensible database, if you will.We would call it a store, because we don’t plan on having the system used for transactioning in our online system.It’s designed to put users, network elements, and printers, and the like in it.And be able to have highly available, yet dynamically changing schema.

In terms of security, there’s a wide number of scenarios that we’ve been asked about, single log-on, private communications for e-mail, secure business transactions, even a secure desktop, and the scenarios go on.There’s a lot of talk in this conference about all the improvements that we’ve been doing in security, but it’s across the board.We’ve added new security capabilities to the system.I didn’t mention in the file system area previously that we’re adding an encrypted file system capability to NTFS. I mentioned IPSEC and ICMP.In terms of authentication, the key area as we move ahead.What we’re trying to do is, do the best of a private key system Kerberos and the best of a public key system using X.509 certificates. Above that, just in terms of safety, you’ve heard about Authenticode, hopefully you’re using it.We’re going to take Authenticode and use it for driver and DLL signing, and actually signing in the system.

The key part of this security slide is that it’s integrated.So, when we talk about the Kerberos keys and we talk about the certificates for the public keys, it’s all stored in the directory system.And these two just fit hand-in-glove together.

There are also a set of other related services that we’re including in Windows NT 5. One is a distributed filing system that we’ve continued to make progress with.The goal here is location independent of the data.It also gives us a way to load balance, and also if one system fails, it’s got a replica.The other system will automatically pick it up.We use this inside of Microsoft today, and it works out incredibly well.It’s also interoperable in a variety of ways with other protocols, NFS, NCP and the like, because you can graft together parts of the names space into this distributed filing name space.So, it’s a way to use the systems that you already have, that customers already have in their sites.

A small thing that you may not think about very much is that you have this distributed system, how do you keep the times synchronized?Well, it turns out there’s been a lot of work, and for some of you it may be redundant for me to say this, but there is — there are specs that describe exactly how you can do this in a standard way.So, we’ve included those sorts of protocols in Windows NT 5.So, it automatically keeps the time synchronized across the network.What’s really cool is that we’ve been able to work with the rest of the industry, and produce specifications for the schema about how they should fit in the directory system.And today we’re announcing that a whole variety of companies have committed to investigate changing their systems to be able to work in this.And so, our dream is to be able to take clients, servers, and everything in the middle, and drive them through the directory service.So you have one central point that you can end up administrating.So we think this is super cool and a big advance for the industry.

In terms of applications services, you heard a lot about it, and hopefully saw a lot from Chris and Cam, in terms of the option pack, and what available at IIS 4, NT Server, the message queue and the like.The whole idea here is to have you avoid writing code, and let you focus on the business problem that you’re trying to do.

There are two areas that I want to spend time on, which is a new installer that we’re going to provide in Windows NT 5.0, and separately.And what we’re doing in the Windows NT 5.0 time frame, to improve on this NTS 4.0 option pack software.

First, in terms of the application installer, make no mistake, it’s a mess.You know, there are shared DLL conflicts everywhere.Per user, per machine, is terribly confused.And de-install is generally just impossible.And that’s an interesting spelling of that.At any rate, if it works it’s a miracle.And people get confused, you know, they go the file system, they delete part of their application, and they think it’s gone.It’s not gone.So there’s a whole series of problems.What are we doing about it?Well, we’re adding the new installer service directly to the operating system.And at the same time, we’re going to have shared components only come out in service packs from us.And we are getting much more efficient in terms of our delivery of those things, and using patch technology the size of those service packs will be much smaller than what they are today.

(Applause.)

MR. ALLCHIN: I thought you might like that one.We also are going to have a much stricter adherence in the Windows logo program, to insure that people use this installer technology.I don’t know how to get out of the mess we’re in, unless the operating system takes over the DLL dependency problem.In terms of this new installer, it’s basically simple.You declare your state, all the things, all your DLLs, all your registry keys, and the like, to this installer service, and then the system takes over the installation and de-installation.Further, it works in a network environment, and you can advertise your application, and we can fault in part of the application across the network, as needed.

Now, the way this installer service works is it runs at an elevated privilege within Windows NT, and so it’s able to put your DLLs and other things in special places.So we’re basically going to lock up parts of the system, but we’ll have a way for you to get into it through this installer service.

The other thing we’re doing is we’re extending the option pack.There isn’t a lot new here, but some pretty interesting things.It will be fully integrated with the DF, for both discovery and security.That right there is a big statement.It’ll make it much simpler to develop the applications.Set up will be integrated directly in with the operating system, today in the option pack there are a lot of different set ups that you have to deal with.It’s certainly separate than thebase operating system setup.That distinction will go away.

Finally, administration will all work through the Microsoft management console.Now, a key thing that we’ve been asked about it, well, you just can let processes run wild, you don’t have any way to control them.So we added in Windows NT 5.0 a new object in the kernel that’s called the job object.And you attach this job object onto a process hierarchy, and you can control what that process hierarchy does.So that gives you a tremendous amount of capability to say, you can only use this much memory, or this process tree can only use this much memory or this much CPU, and then we shoot it in the head.And so what this gives you is the ability on a Web server, if somebody writes an errant application, or decides to suck all the CPU cycle, it is only going to have so long to go.So it’s a much more controlled environment.Also we are focused on performance in this area.

In terms of the user presentation area, we’ve made a lot of progress. We’ve added DirectX 5, so Windows NT 5 will be the development platform.We’ve added feature parity, and common drivers between Windows NT 5 and Windows 98.So for the first time you can get hardware acceleration for 3D and sound, following the direct model, full duplex in direct sound audio, and force feedback joysticks.And if you haven’t tried one, you’ve got to try one, they’re unbelievable.So that’s sort of in the fun area.

In an area that’s much more serious is in an emulation system that we call FX 32.It allows you to do a single binary X86 that runs on Alpha.So you’ll be able to test your 32 bit applications on an Alpha machine, with no changes.It means that ActiveX controls will now run on Alpha with no problem, which is a big, big thing.What we’re going to do is show you something that — you may love Windows NT, but you’ve always wished in your heart it could do one thing, have some fun, and play a game.

So with everything I’ve talked about, particularly in the distributed area, you could say, oh my God, there’s no way that it’s going to be any easier to use and, in fact, it could get a lot worse.Well, while we haven’t really focused on simplicity in this release, we have focused on management.And there’s three things I want to discuss with you, which is intellimirroring, the unified instrumentation capabilities added to the system, and the Microsoft Management Console. Now, Intellimirroring is composed of many pieces. Intellimirroring is really about seamless machine replacement, full roaming user support.So you can move from one machine to another, and all your custom changes and all your documents follow you.Just in time replication of the data and code.So you only get the pieces that you need on your machine.And offline access for both settings, user data and applications.And this is quite complicated.And it’s something where you will have to help us in being able to separate your user state from your app state.H-key current user is a great thing to learn about, if you don’t Know about it already, and that’s a key part of separating these aspects.

And all we’re trying to do with Intellimirroring is to to have one machine totally die, and replace it very quickly.The hard disk on the local machine is used as a cache, so that all the information, including operating system, customization, profile, user data, whatever, flows back to the client.And if you make changes it flows back to the server.And if you’ve got two people manipulating things, then you’ll get conflicts, and the system will have some choices about the automatic resolution of that, tremendously powerful.And because it uses the local disk as cache, it doesn’t end up taking more network bandwidth, so it’s unlike other approaches that force you to be on the network all the time.

In terms of instrumentation, we really haven’t had a very good story.It certainly hasn’t been very comprehensive.So we have something called WEBM based enterprise management, which most of the industry is behind at this point.It’s a way to describe and store all the management data and events that are going on on the system, whether they’re coming up through drivers, or coming up through the registry, or coming from applications.And it also works so that you can access this in a cross-platform way, from UNIX or Windows.So if you’re writing a management console, you should be writing for the WEBM environment.

The other thing that I mentioned earlier was WMI.And what it is, is a new capability inside the driver system, so that you can expose events and other status information dynamically with what’s going on with the drivers.And if you do that, it will automatically be promoted up into the WEBM world.Now, what’s really cool about this, as you might expect in WEBM, there is a schema there for those objects.Well, guess what, it matches the schema environment in the directory system, so there is a nice clean synergy there.

So if Keith can come back up, we’re going to do a replaceable machine.

MR. WHITE: The piece of Intellimirror that we want to show you now is this whole idea of replacing a machine.So you have a computer that you’re working on.For some reason the hard drive should crash, or you replace — or you get a brand new machine, you want to be able to easily transfer that information to the new computer.Again, in the past with the machine crashing, you know, it takes days and days to get all that information back.So here I am on a Windows NT 5.0 client.I’m going to go ahead and turn on the Active Desktop, because it will also save my settings, as well as, if I go to my documents here, you’ll notice that I don’t have anything in the my documents folder.So let’s go ahead and create a Word Pad file.We’ll just say welcome to PDC.We’ll save this.Let’s just call this PDC.And that should save this document into the my personal, or my documents folder.

So again, so we’ve customized the desktop.We’ve saved an application to the local hard drive.And we close this down.And as we’re shutting the machine down, unfortunately —

MR. ALLCHIN: It’s going to crash.I know it.

MR. WHITE: Just go ahead and pull the plug on this thing.And the hard drive crashes.So what we’ll do is we’re going to take and replace this with a brand new PC.So we’ll get this piece of junk out of here.

MR. ALLCHIN: You could start this process in a couple of different ways.Windows NT 5.0 really doesn’t require new hardware to do it.We’re using a NetPC here, because it’s got a boot capable prom on it.We’ll also give out a diskette that can be installed in a wide number of machines that will essentially do the same thing as the boot prong, so you’ll be able to use it on currently installed customer equipment.

MR. WHITE: Exactly.So we’re set up and what I want to do is have it go through the BIOS process.And I’m going to show you that there is actually nothing on this physical hard drive.So we’re actually done an F disk, and gotten rid of all the files on this specific machine.So when it boots up, again, being a NetPC it has a remote boot net card inside of it.So it will try to go up to the network and boot.We’ll cancel that process, so you can see it boot off the local drive. So it’s exiting that process.And it should give us a non- system disk or disk error.So, again, nothing whatsoever on the hard drive.So let’s go ahead and reboot this.And what we’ll do it is it will now remote boot, connect up to the server.And then it will download and configure the operating system accordingly.So while it goes through that process, I’ll go to Jim and then we’ll come back and take a look at this.

MR. ALLCHIN: We’ll come back and check on that shortly.

So I’ve been talking about Windows NT 5.0, both Workstation and Server.I haven’t been exactly clear about which pieces are in which.But, it should be pretty obvious.There is also going to be an enterprise version.And there’s going to be a capability, which we call code name Hydra, which basically remotes the Windows presentation onto Windows environments or non-Windows environments, like terminals and the like.So that’s the basic part of the wave that’s building behind Windows NT 5.0.

There will be releases of every one of our BackOffice applications, taking advantage of the directory, the security, Intellimirroring, and the like, that are sweeping behind this.On top of that, there’s a whole set of suites that we’re providing.All these are building blocks, so that you can write applications on.Now, I want to highlight one application in particular — or suite in particular.It’s up here.It’s the BackOffice Small Business Suite.We haven’t said a lot about that suite, but we think it’s going to be very, very successful.There’s going to be a technical session on that this afternoon. It’s one that you should pay special attention to.We think it will be the high volume runner for Windows NTServer for any systems less than 25 users.It’s a very easy to configure and install.It’s basically got about three screens when you start up.It comes with an integrated database, messaging system, proxy system, Internet capable, and on and on and on, fax, modem pooling, all built into this product.

And unlike where you have to install each of the components.It’s not that way.You type about three screens and that’s it.Further, it automatically gets you your domain name from the Internet.So it’s very custom tailored for small business environments.And if you’re doing accounting applications or anything that might fit into a small business, this is a great opportunity for you.

These other suites will follow along, as the BackOffice applications themselves are released.So when?The 98 beta three will happen in the fourth quarter of this year.The enterprise version of NT 4.0 is this month.The BackOffice Server launch will happen October, it’s actually October 22nd.You’ll find information on it on our Web site.You might consider going to that.The NT 4.0, with the option pack, fourth quarter of this year.Hydra will first go to beta in the fourth quarter of this year.That will be for 4.0 and then there will be another version of that, come 5.0.

The Windows NT 5.0 beta one, it’s now.Well, not exactly now.It’s sort of now.It means that on Friday you’ll receive the CD.In fact, you’ll receive all the CDs.It’s build 1671.We’re running it on all the machines here.And you could say, well, why didn’t we give it to you now?Well, maybe a couple of reasons.Maybe we didn’t finish it until just recently.We also thought we should try to encourage you to stay here and not in your hotel rooms playing with it.That was another good reason.But, anyway, you’ll all receive more CDs than you ever wanted, probably, on Friday morning.I think it will be out in the main lobby area.We’ll be putting 200,000 copies of it on MSDN, which will come out in a few weeks.It will have Windows NT Workstation and NT Server on it.It won’t have other products on that.

Now, I should probably stop here a minute and give you sort of a set of expectations for what’s in Windows NT 5 beta one.There’s a lot that’s not in it yet.You won’t find a lot of Intellimirroring.You know, there’s still migration work that we need to do, from NetWare and from Windows NT 4 into the system.There is actually a reasonable amount from Windows 95 to Windows NT 5 Workstation that you might want to play with.There’s a whole series of other things in the directory area, broader device coverage, and on and on.So there is still a lot to do.This is an early beta.

There’s a lot in the preview directory.Those are the things we didn’t feel so perfect about from the testing.You can go in there and play with them.The main system is pretty solid.The worst problem you’ll probably have, if you have a problem, will be during set up, because we’re still doing coverage in the set up mix.I can say that once it’s up, it should be very solid.We are running it.I run it for a ridiculous amount of time.Our stress coverage, once the system is up, is higher than any of our past systems.So we feel very good about the capability there.And on the server side, we’ve been self- hosted, inside the development organization and the directory system for some time.And we have a trust relationship with the rest of the campus.

When is Windows NT 5 beta two?I don’t know.You’re going to help tell us that, along with the rest of the people who are going to get copies of it.We’re going to wait for some feedback to set the date.It’s just — the fact is that we RTMed it on Saturday night.And we’re just going to have to have a little feedback to know.The next beta version of it, though, will have the Workstation, the Server and the Enterprise version.

So where are we going beyond Windows NT 5?I already said, it’s a work in progress.There’s a whole list of things we have people working on today.Preparing for the future.We are going to do a full 64 bit NT.We have been working on that for some time, both with Intel and Digital.You’re going to hear about COM Plus later in this conference.And you probably heard a little bit about it already from Paul.And there’s a whole series of other things that we’re doing, that I’m not going to spend the time in this conference talking about.But, needless to say, it’s got a great future ahead of it.

But, you can wrap all these together for our vision of beyond NT 5 in two ways.First, we’re going to really focus hard on simplicity.Second, we’re going to focus on the distributed computing infrastructure.We’re not done.We’re just getting going.The result of our dream is a much more intuitive and helpful system, one that has lots of different ways to interact with the users.Speech is certainly one of the key ways.

It needs to be information rich, so that the interaction that you have with the system is one where it’s giving you the right information and not drowning you.Seamless communication, we want it to be like electricity or the telephone.I want the networking stuff to go away.Maintenance free, these systems are incredibly hard to use.Our software is, and frankly, the applications are.We can do a lot to make the systems much more resilient.Working together that’s something that customers are expecting.And if we want to drive up market share in the homes for your applications and our system, we must, in fact, focus to this vision.

Keith, how’s that Intellimirroring replacement looking?

MR. WHITE: It’s looks like it’s almost done.The installation is done.So we’re at the log-in screen.So what I need to do now is log in as myself.And what this will do is rebuild my profile, and then intelligently cache my settings and my data back to the local drive.So it’s going through this process.Again, no user interaction is necessary.So what we should do is go ahead and build the background that we change and it should give us our document into my document folder.So it looks like it gave it back to the background.Let’s go to the my documents directory.And we should see the PDC file when we open that up in Word Pad, we should see the doc that we created earlier with welcome to PDC.So that’s the totally replaceable PC.

MR. ALLCHIN: Great.Thank you, Keith.

In 1992, we held the first Win32 developer conference.There were conferences before that, but it was the first time we talked about Win32.In fact, it’s the first time we showed Windows NT.There were 1,400 people there.We were stunned.Today there’s over 6,200 of you here, and we’re still stunned.

In that first conference there were a few pioneers who decided to bet their work on Windows NT.And by and large those companies and those individuals have done incredibly well.The opportunity before you today, with the beta one, and what’s happening in this wave of Windows NT 5, isn’t dissimilar from what happened before, except there’s even more opportunity now.Windows NT 5 is going to change the way software is written.It’s going to change the way hardware is built.It’s going to change the way networks are constructed.And longer-term, it’s going to change the way users interact with their system.

And while Windows NT 5 is a great opportunity for Microsoft, the real winners will be the pioneers, the third parties like you, who decide to invest and bet on the most advanced operating system we’ve ever created, Windows NT 5.Thank you very much.