Remarks by Paul Maritz
Group Vice President, Platforms Group
Tuesday, May 20, 1997 
New York, NY

MR. MARITZ: Good morning again. I'm going to take over and basically go through some more detail behind our product plans and the strategy that's guiding those plans. 

As you know, our overall strategy that we've been operating on for a number of years now is and has been to build out a scaleable family of operating systems. With Windows 95 and Windows NT, we're well into that strategy and we recently complimented those products with the Windows CE operating system, which is designed to scale down into the consumer electronics space, so you see us offering Windows CE into the hand-held space and recently, with our announced plans to acquire Web TV, you can expect Windows CE to go down into the set top box arena as well, so a tremendous range of scalability here.

Today we're talking about scaling towards the upper end of the scale and in particular, scalability with Windows NT servers. Our strategy there again is to span for the server a very broad range, approximately 3 orders of magnitude from a price performance point of view, all the way from machines costing $2,000 for low end servers all the way up to hundreds of thousands of dollars and millions of dollars and the kind of performance that goes with that, as you saw with the demonstrations that Bill did earlier on. Now, the challenge is not just the scale. We're not just trying to achieve high performance, high throughput here. What we're really trying to do is to not only reach using the PC infrastructure model to the most demanding applications, we're trying to do some more important things as well.

We're trying to make sure that we provide a common infrastructure between small, medium and large businesses. If you think about one of the great implications of the internet is that we're going to be able to have commerce and information flow across organizational lines, we would like to make it easier to do that rather than more difficult.

Small businesses, medium size businesses have many of the same complexity issues that a large business faces, just on a different scale, so if we can give them the same infrastructure, we believe it's going to be easier for businesses to do business with each other.

So the scalability is not just being able to reach those high-end demanding applications, but also enable greater flow of information, particularly across organizational lines, and get this type of rich information infrastructure into the hands in particular of medium sized businesses, as well as large businesses.

We're also building out a broader infrastructure, not just what you traditionally thought of as the operating system, to allow more useful usage of information. So a key part of what we're doing is enriching the underlying infrastructure to provide more services that can be used to build more flexible and more useful applications and I'll come through this in a little more detail in a few minutes.

So when you look at scalability, you've got to look at several things. You can measure scalability in terms of the number of transactions you can do on a per second or per minute or even per day basis. You can look at the amount of data that you're handling, in terms of what you're storing and operating those transactions against. You can look at the numbers of users that you're supporting in terms of concurrent active access to that information. 

But you also need to look at other aspects as well, which is the cost or price of those solutions, the price in terms of the cost of the underlying hardware and software and then the cost of operations, what does it cost to actually put all of that infrastructure together and we're focused on tackling every one of these problems.

To do so, we have to work in close partnership with many others in the industry because reaching an effective solution requires not just that you have rich system software there, but that you have the underlying hardware that has to be structured accordingly, the applications need to obviously expose that underlying capability and then you need the utilities and tools that make the whole manageable on an ongoing basis.

So we're focused not only in each of those dimensions of scalability, but also building up the partnerships that are necessary to really realize this for our customers.

So our strategy then is to make a series of important technological investments. We're building up our operating system to handle larger and larger single nodes. We're then aggregating those nodes together into clusters. We're working with others ourselves, with others in the industry to infuse into that platform the necessary tools and utilities to manage the overall solution, and then layer on top of that this richer infrastructure for building applications.

But we're doing so with a particular philosophy in mind that we think separates our efforts from some of the other comparable efforts in the industry. First of all, we have a rigid philosophy of maintaining real symmetry between what's on the server and what's on the client.

In other words, everything that will run on the server can in principle run on the clients as well. That's very important, as you'll see later on, for addressing the fact that in the real world, users are not always connected to the network. In a lot of cases they have to work off-line, in which case you want to have a lot of the infrastructure done on the client machine as well.

We're focused on hiding complexities from the users so that we can get the solution to scale, not only upwards but also to reach the needs of small and medium sized businesses as well. So a key thing of what we're trying to do here is to retain the functionality, to provide a simpler solution that can be supported across very large numbers of businesses.

Then obviously, of course, we must make sure that the underlying cost of the solution remains very reachable, affordable for that broad range of users.
So we have a philosophy of building high volume, lower cost components that can be scaled across a very board range of the industry.

If you go back to the basis of what we're building on here, the real foundation is the Windows NT operating system. As I said, we actually started to develop this operating system in 1988 and at that time frame, we laid out the goals that we were going to build something that was going to be both a file and print server and an application server.

There was constant tension, should we just focus on Novell and be just a file print server, should we just focus on the UNIX market and just be an application server and we decided in fact to be both.

It was in some respects an expensive decision because it took us longer to develop that operating system and longer to bring it to full market fruition than we initially thought. But we stuck with that basic philosophy, which is very useful because when the internet came along, web service came along as a phenomenon, and really combined the attributes which you find in the file and print server in terms of very high I/O throughput rates, as well as the need to also be an applications server, to be able to run those scripts up on the server.

So when the Internet happened, from a server perspective it was a snap for us to basically readapt the underlying mechanisms that we built into Windows NT to address that market.

But it also put into Windows NT the hooks that we knew would allow us, as the hardware matured and scaled right up, even into data center operations, to be symmetric multi process enabled. We made allowance for cost of processing in that system and the result is, as you see, of all of the important server environments today, Windows NT is by far and away the one that is most capable of scaling up and down that range, all the way from file and print services up through web services, applications, database services, eventually into the data center space.

So to do that, we had to put in place a key operating system foundation and recruit the people to do that. If you think back to the late 1980s, we were then faced with the challenge of not only how do we lay a new operating system foundation down, but how do we make the transition from what was then a 16 bit world, as well as the PC was concerned.

And rather than compromise the foundation that we were laying down of Windows NT, we elected to execute a dual-member family strategy. We elected to have the investments we were making in Windows NT and in Windows 95, two complimentary, parallel code bases, one focused on the problem on transitioning off the 16 bit base, the other focused on the problem of laying down this robust scalability foundation for the future.

We went out and recruited a great world-class operating system team to pull off that strategy. We recruited Dave Cutler, who had years and years of experience working in the VMS operating system and other operating systems and we put him together with a diverse team who had expertise in the UNIX space, in the OS/2 space and other operating systems, and they were focused on basically staying true to the fundamental principles that we laid down, which were to not compromise building a robust, secure, scalable, portable platform.

That was the given. If it took us more memory to do that, so be it. If it took us longer to do that, so be it. We also stuck true to the principle of symmetry between client and server, so we built Windows NT for the workstation environment and the server environment.

Then we realized that we needed to really build a distributed system and we needed to have further expertise in our system software development area to really build out what would become a complete networking platform, which is again an important thing for us to have done because when we got into the internet space, we had a lot of the thinking and the technologies ready to take advantage of that.

So we went out and added to our team people like Jim Allchin who is the founder of Banyan Systems, Rick Rashid who developed the MACH Operating System at Carnegie-Melon University and others and they joined our overall team and contributed to that pool of expertise.

The result being that we've now been through four major releases of the Windows NT system. As I said, it's grown from about -- if I remember correctly, about 7 million lines of code when we did our first release back in 1993, to approximately 19 million lines of code today.

There's been tremendous functionability infused into that system over the intervening four years and we've basically built on that foundation, adding to its performance and scalability and functionality all the way up until today, where we're now working on Version 5, which will beta later this year and then ship next year.

From a performance perspective, you've seen our performance go up steadily. This is a metric showing our performance on today's industry TPC-C benchmark where we've been able to raise our performance in conjunction with a single server database by using again performance improvements in the system, additional improvements in our symmetric multi-processing support, where our performance has gone up virtually by a factor of almost four since we started this effort over -- we first measured on the TPC-C benchmarks approximately two years ago.
So tremendous improvements in scaling and we expect to see this continue, taking another big step upwards as we kick in cluster support.

Also, as Bill announced, we're responding to the needs of applying this technology to more demanding server environments, to more demanding application scenarios by further specializing the Windows family.

We will now have what we call enterprise versions of our server offering, so Windows NT Server will now come in a standard version and an enterprise version.
This enterprise version of Windows NT Server will actually be available in the third quarter of this year, 1997, based upon the NT 4.0 code base, so we're adding key functionality into that code base, supporting eight-way symmetric multi processing out of the box, pushing that way above eight-way in conjunction with hardware partners.

You saw an announcement last week in conjunction with Tandem where they ran a benchmark an a 64 processor system running Windows NT, adding in the cost of support, and as we work toward Version 5.0 of Windows NT Server, the Enterprise Edition will add in 64 bit support initially on the Alpha platform and then in the future, on the Intel platform. So what I'd like to do now is to invite Tanya van Dam who is a product manager at Microsoft, to come up and show us that underlying performance in the guise of an internet web server.
(Demonstration)

MR. MARITZ: As I said earlier, having an operating system is not enough, just in terms of being able to do file I/O and basic operations. You've got to provide that infrastructure for applications because applications, after all, are the things that really make information useful to people.

So in having built up our base operating system team, having built up our distributed networking team, a couple years ago we started to make a very serious investment in building literally a world-class team in terms of distributed applications infrastructure, database technology, transaction technology.

And we put together a great team to build the distributed applications infrastructure, some of best people in the world, including people like David Vaskevitch, Hal Berenson and of course, Jim Gray, who was up on stage earlier with Bill and I'm sure Jim will forgive me if I say when it comes to database technology, they don't come any heavier than Jim. That's a figurative statement, of course.

So we put that team to work and for the last several years they've been guiding our investments in the strategy that we call our BackOffice strategy, which is to put together not just a single application or individual applications in the server environment, but to put those together into a common underlying platform that runs on the server, that additional distributed applications can be built on.

The key investments in the BackOffice environment have clearly been investments in SQL Server. A couple years ago, we took over the database work from Sybase and we started to really invest very heavily in that technology.
You'll see later on that it's not just an idea of being in the database business just for the sake of being in the database business. We're working in this because there's key technology here that needs to get integrated.

The next key investment is obviously building up an industrial strength scalable messaging and collaboration server environment, which is our investment in the Exchange Server and then recently we've complimented those two key members of the BackOffice family with a third offering that we call our Site Server product, which is designed to provide you with a lot of the analysis and management tools that you need to really run an Internet site.
So these three investments really contribute, from a product perspective, the key investments that we're making in the BackOffice family. I'll come back and give you a little more detail into the technical ideas behind these investments in just a minute or two.

But before I do that, I want to give you a demonstration of Exchange scalability. Exchange has been a very important investment for us because we had to build a messaging and collaboration server that would be a new client-server architecture but would be a step upwards for the millions of customers that we have running Microsoft Mail.

So this is a product that we knew had to hit the streets running in scale and be robust from day one and I'm happy to say that we really have been able to achieve that and we have some very, very large customers indeed, all pushing now close to the 50 thousand users mark already deployed in the environment and on their way to deploying Exchange service supporting over 100,000 users, as you saw, for instance, at Lockheed in the video earlier.

So again, what we actually have running upstairs is a copy of the upcoming release of the Exchange server that will ship in the second half of this year, which further extends Exchange scalability, in particular extends it in the direction of handling large amounts of data.

So what have you got to show us here, Tanya?

(Demonstration)

MR. MARITZ: So as I was saying, we're continuing this theme of building up a standard and enterprise version of our key server products, so just as we're doing a standard version of Windows NT Server and an enterprise version of Windows NT Server, we'll be doing Enterprise Editions of our BackOffice suite of products, so both the suite itself and the individual constituents of that suite.

In addition to scaling up, it's very important, as I said, to scale this architecture down to be able to reach small and medium size businesses, so I do want to note that in the summer of this year we'll be releasing a Small Business Edition of BackOffice where we've gone through and dramatically simplified everything to do with the BackOffice environment.

The challenge was to get it down to basically where a small business could just take a CD, answer a few questions as to what the name of their business was and the name of their users were and basically have the whole system automatically install itself and provide them with a server operating system, a messaging environment, the database environment, the fax environment, et cetera, as well as a server environment and a proxy, all of that integrated into one package.

We're complementing that, to return to the theme, though, with the Enterprise Edition of BackOffice, which is comprised of the Enterprise Edition of the key members of the BackOffice family, which includes now an Enterprise Edition of SQL Server, so later this year we'll come up with the Enterprise Edition of SQL Server 6.5, which offers increased performance, it actually takes advantage of additional memory space that we're freeing up in NT Server Enterprise Edition and very importantly, exploits the cluster environment.

We're coming up with Enterprise Editions of Exchange, extending the database message size of Exchange and again being able to exploit the cluster environment, and likewise, Enterprise Editions of our Site Server product that will particularly add in the electronic commerce environment that we have, our Commerce Server product that we're building up will be added into the Enterprise Edition of Site Server, with the very high end usage tools that are necessary to really analyze what's happening on your commerce sites.

So to demonstrate those Enterprise Editions, I'm going to ask Tanya to demonstrate to us the upcoming releases of NT Server and SQL Server, so what we're going to be looking at is NT Server 5.0 and the upcoming release of SQL Server, which are in particular, as I said earlier, going to be enabled to handle very large amounts of memory.

What we're demonstrating here again is an Alpha server, which has 64 bit addressing in it so that Windows NT Server 5.0 can take advantage of that memory and expose that memory in turn through to the SQL Server database.

(Demonstration)

MR. MARITZ: I wanted to talk a little bit about the work that we're doing to provide this richer set of infrastructure for applications, particularly distributed applications. The real challenge here is as we try and move businesses to the next level of productivity, we have to obviously automate more deeply the processes that underlie those businesses.

If you look at almost any business process today, you have to combine both structured and unstructured information. If you take any example from a customer service environment, if you want to find out everything that's relevant to a particular customer, you might have to look at billing records that are in some structured database, you might have to look at e-mail messages that have come from that customer that are sitting in the message database, you might have to look at documents that have been written to that customer or policies that have been stored as documents.

So we need to be able to have these processes run over different sources of information. The problem is today that most sources of information sit inside their own application's infrastructure. They have their own way of dealing with queries, they have their own way of dealing with security, they have their own way of dealing with scripting and customization.

This makes it very difficult for businesses to write processes that cover these different sources of information. Typically, inside a business day you'll find a whole infrastructure dedicated to their classical line of business, record oriented applications. You'll find another infrastructure that's been put in place to handle work group or messaging.

You'll typically find a third or a fourth infrastructure that's been put in place to handle their web site environment. It's very difficult to pull all those together and treat them as essentially different facets of the same underlying problem of dealing with information.

So what we've been trying to do is not only provide great instances of each of those four important classes of stores and have them as part of our underlying platform, either part of Windows NT Server itself or part of the BackOffice platform in terms of a great implementation of a file and a document store, great implementation of a web page store, great implementation of a message or work group store, great implementation of a database store.

So we've not only been trying to do those and provide best or great examples in every one area, but we've also been systematically going through each of those environments and saying what can and should be common about those systems.
So if you look at an example, for instance Transaction Server, we elected not to build that Transaction Server several years ago when we started that investment, as part of the database environment. We deliberately abstracted that out and built it as a separate service that could run across these other stores, so that's now integrated into the underlying operating system and is now available in the future to put transactions behind processes that might, for instance, involve the sending and receiving of electronic mail messages, so if you have some important process that requires a mail message to be delivered in a certain time or read within a certain time, you'll now be able to put that into the same idea of putting it under transaction control and being able to treat it as an important step in that process.

We've also been going through systematically isolating the scripting and customization environment, which is typically part of the web server environment, pulling that out, making it common across these other stores.
The next step is to pull the query environment out from the database and make that query processor available across all these other stores. You can issue that query that says get me all the information with respect to a customer, whether it be records in the billing database or messages in the message store or documents in the file system.

So that team of architects that we've recruited for the distributed applications phase have been working at laying down the common infrastructure to support these distributed applications and that infrastructure is becoming part of the Windows NT operating system environment. Not only on the server, but in the future on the client as well.

That's a very important aspect, because if what we're really building towards is the ability for an individual to get that customized feed of information down to his desktop and the customization is being done by the infrastructure on the server, what happens when that individual disconnects, basically, on his laptop, because a third of all PCs in business today are laptops.

When that happens, when that person disconnects, you want to, speaking figuratively, have the server come down onto the laptop and live on the laptop for the duration of the period that he's disconnected and when he reconnects, have everything sync up again.

So that symmetry between Windows NT Server and Windows NT Workstation is a critical bedrock that we can use in the future to continue to provide you with rich access to information, whether you're connected or not connected in the future.

In addition to our building out the operating system, the applications infrastructure, we need to build the management support into the system, so we've been working hard at building into our platform compatible protocols and networking infrastructure.

Several years ago we were fortunate enough to select TCP/IP as our key protocol and build that in a high quality way into our operating systems. We've continued that theme of building out the plumbing that we build into Windows in an Internet compatible way.

Our vision is that the internal network that you build out should as seamlessly as possible integrate with this new public network that's emerging with the Internet. So you get in Windows, built into the environment, Internet-ready networking as part of that environment.

We also provide the important legacy networking environment, particularly the SNA environment by virtue of our SNA Server product and we also continue to invest in the Systems Management Server, which is designed to allow you to administer a large network of servers and PCs.

Continuing that management theme, as Bill mentioned earlier, there are very important investments that we're making, most important of which is our Zero Administration Windows initiative that really targets total cost of ownership, really allows us to take a fundamental bite out of the things that cause people to have to spend money today on owning and operating distributed networks of PCs. 

As part of that overall initiative, there are certain management features that are being built into the Windows NT environment, directory support, common console support, and we work very hard to make sure that these facilities that we build into our platform inter-operate with the solutions that have been provided by the community of enterprise networking management vendors.

We've also cooperated with the broader community on defining new standards for management. Basically, specifying a schema that defines the type of information that can be shared between management systems and this is a very important initiative that we call our WBEM initiative or "web-em" initiative, which is now being ratified as a management schema by the Desktop Management Taskforce.

What we've been building out is a platform, a Windows NT/BackOffice platform which is designed to run applications that make information more available to users either directly or in conjunction with other applications that get layered on top of that platform.

So we view this as a platform that we evangelize to the industry at large and we try to work as closely as we can with the key application vendors so that they get full benefit of this platform and we're seeing some rewarding results there as well.

For some years now we've been working very closely with SAP who have moved their R/3, which is the client-server enterprise application, onto the Windows NT platform and SQL Server platform. The result is today that about a quarter of all R/3 implementations are on the NT platform and on an ongoing run rate, in terms of new systems being sold, almost half of R/3 installations today, about 42 percent, are on the NT platform.

And they're also working very closely with us to make sure that as we expose those new features of the platform, that they take advantage of them in their application.

Recently I concluded an agreement with PeopleSoft where they've licensed 1,200 copies of the BackOffice environment for their internal use. It will be their standard server environment, their standard platform on which they build the applications.

If you look at Windows NT itself, you're seeing support from other vendors there, such as Oracle recently, last week, demonstrating the Oracle server now running on the NT environment and as I mentioned earlier, last week Tandem through their NonStop/SQL environment demonstrated a 64 processor cluster system, handling a database with a huge number of rows in it, a very high end demonstration of scalability indeed.

So you're never done with scalability. It's something that you continue to work at forever. As I said, for the last eight years we've been working on this platform, we've been through four major releases of Windows NT and we're working on the next release. We have the release after that in planning, so it's something that becomes a way of life that you just work at, so we have continuing work to make sure that as the hardware improves underneath us, that the software doesn't become the bottleneck, that we have the necessary support in there to expose our customers to the miracle of semiconductor revolution and the networking revolution.

We're working hard on simplifying and extending the distributed applications infrastructure, making it ever simpler to build these rich information applications.

And lastly, but not least, it's very important, working on the management infrastructure, building in the facilities that you're going to need to build an automated, complete management environment.

The result of that scalability is we'll return to those themes of not only enabling scalability, not only being able to take the PC architectures and expose their benefits to the very highest end applications, but also in doing so, open up the barriers between businesses, provide a common framework, a common set of conventions that can be relied upon between small, medium and large businesses.

And also really make information more useful by enabling a new generation of more flexible applications. With that, I'm going to close up and invite Deborah Willingham to come up and join me. The chairman of SAP finally managed to land, that's the good news. The bad news is he landed in Hartford, Connecticut and he's been driving down here. I hope he isn't driving and driving at German autobahn speeds. We'll pay his speeding tickets, however. He's going to, we're told, be here, but he'll speak after Deborah. So I'd like to invite Deborah Willingham, who is our Vice President of the Enterprise Customer Unit, to come up and join us.