Keynote Transcript


Sapphire '97

Andrew S. Grove
Orlando, Florida, USA
August 8, 1997

JEREMY COOTE: And so it is with great pleasure I introduce Dr. Andy Grove, the Chief Executive Officer of Intel Corporation.

[Music and Applause]

ANDY GROVE: Hi, Jeremy.

JEREMY: Good morning, Andy. It's a great pleasure to have you here today, not only as a keynote speaker this morning but also as a CEO representing one of our key customers.

ANDY: I'm very disappointed that you didn't ask me to dance.

JEREMY: Well some people are wondering why you didn't come in a rabbit suit. They were hoping you would.

ANDY: It's not a rabbit suit - it's called a "bunny suit."

JEREMY: Oh, I see, it's called a bunny suit. So, we'll just take minute to get the equipment set up that you will be working with. I think it will be a great session this morning and we all look forward to it.

ANDY: Thank you and good morning. In 1993, a a relatively disheveled but also very excited Hasso Plattner visited me for the first time at Intel Corporation. He came in very excited and talked to me about something that I - at first - didn't quite know what it was all about. It turns out, that just before his visit to Intel, which was the first time that he and I had ever met, he had seen a demonstration of SAP software running on the first Pentium® processor-based server. He was very excited about the demonstration and mentioned to me that he had just seen the beginning of a new era. The only thing that marred his visit was that he reached into his pocket and handed me a letter in which Intel had informed him - that for a variety of reasons, we disqualified him as a bidder for our enterprise software. That was four years ago - which is a lot of time. Today, I'm going to give you an update on the progress that we have made since then with Intel Architecture-based servers. And since then, Intel has also become a major SAP customer which I'll talk more about as we go along. But Hasso's visit - in a sense - could be described as a sign of a strategic inflection point; the beginning of a new way of doing business.

Let me start by talking a little bit about what strategic inflection points are which is the gist of my book, SAP was kind enough to give each of you a copy of. Beyond that I'll talk about the current trends that take place in business computing.

To start talking about strategic inflection point, I'd like to review very briefly the framework for competitive analysis from about two decades ago to the work of Michael Porter at Harvard Business School. This work describes the various forces that affect a business and keep its competitive well-being in a particular stage. The question is what happens when one of those forces becomes particularly large, let's say for example the ten-fold magnitude. For instance, what happens if the way of doing something can be done in a differently and becomes ten times more effectively? I have sketched out a diagram that describes what tends to happen there. The shape of a business, the competitive well-being of a business, goes through a subtle change that I call the inflection point. Depending on the action that the firm takes at that inflection point, the competitive health of that business can go on an upswing or can begin to decline. And although these two curves merge very closely at the strategic inflection point, you can hardly tell which direction you're going to be heading after that. That's the nature of an inflection point. Once you begin your ascent or descent, the course that you've committed to tend to be very difficult to reverse. That's what strategic inflection points are about. They can be brought about by a variety of things, typically a major change in the competitive scene. Technological change is one. Regulatory change, for instance, when the telecommunication industry in the United States was deregulated first in 1984, caused a strategic inflection point in the telecommunications industry. Changes in market perceptions and demands which are very subtle and very hard to pinpoint, changes in fashion could be one. One characteristic of the changes in market perceptions is that that quite often those of us in senior management are the most oblivious to these subtle changes, and the people in the field, the people in the front lines, tend to notice those earlier and they start responding to that. Under those circumstances, the strategy of the corporation goes one way and the action of the corporation goes another way causing what I call a dissonance. And in all of these instances to cope with the strategic inflection point and catch that upward curving portion of the curvature requires a fundamental change in a company's business strategy and generally a change in the core competencies of the business itself.

An example of a strategic inflection point in the computer industry is the advent of microprocessor-based personal computing. Many of you have grown up in a computing industry that was characterized by a vertical alignment of companies. Each company created its own hardware platform, typically generating its own operating system, their own applications and sold their products through a direct sales force. The means of competition in this business was between the vertical columns. This is how competition in the industry was waged for the first three decades of the computing industry - until a ten-x force came in the form of very cost effective personal computing that leveled all of this. The vertical industry was turned into a horizontally based industry which meant several choices of microprocessors, several choices of operating systems, software applications coming in packet form. Equally importantly was for this horizontally arranged industry was that the distribution of these computers came through a heterogeneous indirect sales force or channel.

Let me tell you a little bit about the strategic inflection point whose first glimpse at Intel was initiated through Hasso's visit which I described earlier. Our information system environment was a legacy of the vertical industry, relatively dated. It was based on proprietary software, requiring hundreds of software engineers to update and modify it. Every change in our business environment was reflected in requests for software modifications that ended up sitting in a queue for many months, sometimes many years. In addition, anyone who used this software required very specific training to understand the proprietary legacy system that we used. While we struggled with the legacy software, Intel continued to grow - we've grown year after year at a rate of approximately 30 and 40 percent compounded annual growth rate. But the information system did not grow in accordance. Instead, the system was patched and patched and patched, until it finally required a fundamental overhaul. So around that same time as Hasso's visit in 1993, we decided to standardize our business on SAP's R3. In the spirit of the "new" horizontal computing industry, we based our utilization of software, on the work of many thousands of people who have been devoted to formulating, developing, and embellishing system software and bring many improvements for our business processes that we could identify in our own business processes.

Equally importantly, from our standpoint, was that we could use this software for other areas of our business, such as enterprise sales order entry and various other business processes. It was also the first time that it allowed us to use one of our own products, our own microprocessor-based platform product at the client level, at the application-server level, and at the database-server level in a heterogeneous software environment. This system, after being up and running for several years, was capable of supporting between two and three thousand users. It has represented the expansion of the strategic inflection point that started with packaged application software, word processors and spreadsheets fifteen years earlier. It has extended up into the most mission critical of Intel's big-time mission-critical enterprise applications which is the taking and processing of orders.

This is how it [strategic inflection points] works. In the process we have had to change our core competencies throughout the company both in terms of the information technology, organization, and in terms of the user community. We are not unique and what we continue to see, are the current large trends in business computing, similar waves are washing through information technology organizations and the companies that they serve and computing users within these companies.

So let me start by concentrating on the first mega-trend which is the growth of the high volume servers based on the horizontal computing industry. These personal computing systems are based on standard microprocessors and standard operating systems. And the foundation for this is a fantastic volume economy that is driven by the PC industry. This volume growth has gone on for fifteen years now and it is flirting with reaching an annual volume of hundred million units. It is a very, very cost effective industry taking advantage of not just the volume economics but also the various benefits of this horizontal industry. Competition on each of those levels brings about an unprecedented continued increase in computing power--which is the blue chart --and an even more impressive drop in the cost per computing power which every two years seems to decline to less than thirty percent, sometimes twenty percent of what it was two years previous.

What is important about this first mega-trend is that the volume economics derived from the PC industry, now also applies to servers as well in the form of what we call standard high volume servers, also abbreviated to "SHV." The same economics that brought this enormous cost effectiveness to desktop devices is bringing it to the server industry. Not only is volume growing and not only is price/performance improving but the features and the design of the standard high volume server are evolving as well. Before the SHV specifications have come into existence--as the SHV initiative has come into existence, servers were largely desktop PC's turned on their side, typically involving one maybe two processors. It was a major improvement that they started to use the PCI internal bus system. The current state of the art SHV server system represents one to four processors per system operating in a symmetric multi-processing mode, with 512K L2 cache, 4 gigabytes memory and an expanded version of the PCI bus.

But equally importantly, the volumes for servers are growing, driven by the same type of dynamics, as within the PC industry, actually at an even faster rate than the desktop volumes. Many at Intel believe that in a few years time we will be looking at quite a healthy number of million units per year industry of high performance standard high volume servers. The same way that this phenomenon brings cost performance benefits to the desktop industry, it has already brought very comparable declines in cost performance on standard high volume servers measured, in this case, dollars per TPMC, as demonstrated by today's announcement of a six-wayIntel Architecture-based server from Unisys reaching the $39 per TPMC level.

This is where we are today. The dynamic efforts of an entire have brought us to where we are today and will continue to take today's status and bring it forward ever more effectively. Many of the factors that drive these efforts start within my home territory, the Intel microprocessor road map.

I'd like to talk a little more about our improvements in I/O, improvements in clustering technology, and improvements in scaleability.

What we at Intel do for a living is build high-performance microprocessors, and as a rule of thumb, deliver that faster microprocessor at the same unit cost as we delivered a slower one a couple of years ago. What you see here is a 32-bit line as well as several "yet-to-be-announced" products. These products will deliver continued improvements in price/performance all the way into the next decade.

The last processor on this chart which has been named is the Pentium II processor that we've introduced earlier this year. It is a truly outstanding architectural improvement that is ramping into multi-million units per quarter volume production today. And we are delivering this capability to the industry in desktop systems, notebook machines, and server systems, bringing the same cost effectiveness but at a level of performance that is really unprecedented by all previous Intel microprocessors. This level of performance also requires fundamental modifications in the architecture of the microprocessor microprocessor as well as and in the rest of the system in order to avoid the system choking down the processor. Taking all aspects of the system into consideration will allow us to remove this bottleneck and let the processor performance actually reach the application and the user.

What we are talking about here, is the limitation of the system bus that is required to access the level 2 cache, which is very important particularly in server-based applications. A new bus architecture--what I will call a Dual Independent Bus architecture--splits the access to the level 2 cache that you see marked as cache-bus from the system bus, allowing 95% of the processor to memory transaction that takes place between the level two cache and the processor to be unfettered by the system bus limitation. In addition to that, we are capable of scaling with increased processor speed rather than having to be clipped by the speed of the system bus. If you look at the internal system bus, a measure of bandwidth in the system bus would be given by this formula.

Let's take a look at how this works in the dual independent bus architecture. In this example that we have the Pentium II processor would operate at 300 megahertz providing 2.4 gigabytes per second of bandwidth. Added to that the system bus performance at today's 66 megahertz bus performance, adds up to a total of almost three gigabytes per second of bandwidth. So when we compare the impact of the Dual Independent Bus architecture with older bus architecture, [the older bus architecture that by and large ends with the Pentium processor is shown in the green]. The Dual Independent Bus architecture with the first physical implementation of what we are calling Slot One, used predominantly for desktop machines, is substantially higher and continues to scale with system bus frequency and with the first of those new processors that I showed on the chart that we will be introducing in the first half of '98. We will introduce a new connection scheme that will almost double the bus bandwidth As you might have anticipated, we will call that Slot Two. So we have removed, in a very fundamental way, bus limitations. Microprocessor speeds will in fact reach applications that translate into to user accessibility – users will benefit in terms of better cost/performance, when the internal limitations have been removed.

Microprocessor developments also continue on in another dimension. We will be introducing the first IA-64 processor code-named Merced, sometime in the next couple of years. And that will be the beginning of a parallel course of developments. Two sets of teams working on the 32-bit product lines and further extension to the 64-bit line beyond Merced. I'm happy to announce today that SAP has elected to become a Merced development and definition partner so by the time the Merced processor-based systems appear in the marketplace, we will have well-tuned SAP software appearing on the scene as well. A 64-bit processor architectures clearly of great importance in on-line transaction processing, which is the performance level we expect to gain from the IA-64, 64-bit processors, starting with Merced.

So that's what's coming in the way of processors. Processor performance is only as useful as the I/O allows it to be. And we've talked about the internal I/O in the motherboard and I'll be able to go further in looking at the I/O between subsystems in a server. The major development that we have been involved with our partners in the industry is the intelligent I/O or I2O architecture specification. This effort has the support of dozens of players in the industry who broadly accept it. Significant here is the uploading of I/O interrupts from the host processor and the operating system allowing subsystems to communicate directly with each other without having to make operating system calls thereby giving a real boost to performance particularly in configurations where there is a substantial number of clients. Similarly we are working on connection schemes between computers to allow the scaling computing power by clustering of different SHV's. What we are working on here is the Systems Area Network for a virtual interface or SAN VI Architecture. This is also an industry specification involving a number of key companies in the industry. This initiative was announced in April together with Compaq and Microsoft.The specification is expected to be finalized in the latter part of the year. It will allow a standardization of the way clusters of SHV's, standard high volume servers, communicate with each other and bypass system bottlenecks.

I feel very privileged to be able to show you the workings of the first demonstration of a SAN VI Architecture compatible set of Compaq ProLiant servers. Let's turn to Annie Lung and let Annie demonstrate this machine. Hi Annie.

ANNIE: Hello, Andy.

ANDY: Maybe you can start by explaining what we have here.

ANNIE: Sure. On the screen as you can see and also displayed on this monitor, we have here a rack mount system consisting of four Compaq ProLiant* 6500 servers.

ANDY: One, two, three, four.

ANNIE: Right. And this screen just shows one of the screens from one of the servers that we're going to use to drive the demonstration today. This technology demonstration also uses the Tandem ServerNet* cluster interconnect in relation to the VI API. The VI API is running in emulation mode using the server processors.

ANDY: I guess it has to run in emulation mode because the hardware doesn't exist just as the specification don't quite exist.

ANNIE: Right. We don't have it today. So what I'm going to show next here is the static chart.The red bars are showing the typical performance latencies of communications of TCP/IP over 100 megabits ethernet connection. So for example, at your end, we're showing 256 bytes message data size, transferring at 200 milliseconds transfer rate. And here, the latency here for 4 Kbyte is over 1000 microseconds. So VI will improve the numbers by a great deal because of the performance improvement and increased bandwidth as well as reducing the latency. Shall I start the demonstration?

Okay. What the system's doing right now is sending a message across all four nodes. And we're capturing the measured response latency dynamically and charting it as you see them on the screen right now. So as you can see, for a 256 byte message size, we are getting approximately half the size of what a typical TCP/IP latency would be and for the 4 Kbyte size we're getting a significantly reduced number as well.

ANDY: But I think it's a key point to stress that this is only an emulation so when the real hardware exists, we're expecting to see an even better performance than this.

ANNIE: Right. So what we have here in the next slide is showing that what we are estimating the performance enhancement will be with the native VI Architecture hardware when it becomes available next year. We're going to be estimating about six microseconds for 256 bytes of message size and---

ANDY: Six microseconds versus the current 200.

ANNIE: Well, the current 200's using the VI emulation but much higher. You're right. 200 using TCP/IP, much lower. Okay? Great. And just to point out one thing, with VI architecture especially for OLTP applications such as SAP, the reduced latency improvement will be a very important asset, and will reduce the transaction processing time as well as reducing the data contention for the database.

ANDY: Thanks, Annie.

[Applause]

ANDY: Now scaleability is a very important factor for companies as they grow. It's a very important consideration for us at Intel as users of information technology capabilities. And scaleability is something that is mobilizing a lot of forces in this industry. Again, a key building block for a larger system from our standpoint was a microprocessor that increases the internal L2 cache size. Just a couple of weeks ago, we announced the first-- the Pentium Pro processor with a one meg internal Level 2 cache that offers the highest performance possible with transaction processing applications particularly in an SHV configuration. What transaction scaleability gives us here is best seen if we look back in terms of history of Intel Architecture based server evolution going all the way back to the four-way Pentium processor that processed 36 thousand line items--sales and marketing data line items per hour. As we move to the four-way Pentium Pro processor-based system, this number jumped up to 112K and today there was an announcement by Data General of a six-way machine that brings this to 145 thousand transactions per hour. That is still the status today and by no means an indication of where we're going to go in the future. Where we're going in the future I would like to give a glimpse of by calling on Annie to demonstrate to us again a first ever 8-way Pentium Pro processor-based system from NCR. Good to see you again, Annie.

ANNIE: Yes. So what we have here, Andy, is as you mentioned the first 8-way SMP system from NCR and we're running Windows NT* and SAP R3 just to show the audience so they don't have to take my word for it. On the screen you see the standard Windows NT diagnostic. There is a line up there, the second line that you can see the number of processors. It says eight.

ANDY: There's supposed to be a drum roll here with that being a first.

ANNIE: Oh, great. All right. So what I'm going to do here is to do a simple line item entry for a typical sales order that an SAP user might use.

ANDY: This kind of sales order entry is exactly of the type that the benchmarks use to arrange this hundred, 12 thousand, a hundred thirty thousand per hour level, right?

ANNIE: Exactly. So all I'm doing is entering a few items that a customer might want to be ordering on the database. And obviously I can't show that many line items transactions because there's only one system here. Oops, I put in the-- One more zero. So basically, now the order has been processed and all I really need to is save that purchase order that I just entered.

ANDY: Super. So this is sign of live demo here. Eight-way Pentium Pro processor-based system, running a Pentium-processor based client here. This is the client in here. We are told that this 8-way system from NCR reaches a benchmark of hundred and sixty to a hundred and seventy thousand transactions per hour level. So thanks, Annie.

[Applause]

ANDY: Just to sum up the evolution of the developments of SHV servers. The modern SHV server as you have seen can range now from one to eight Pentium Pro processors per server. We have reached the one megabyte level two cache level. It is capable of dealing with sixteen gigabytes of memory. We have multiple PCI buses, intelligent I/O is built into all of these as part of the specification. We are at the verge of introducing SAN VI Architecture clustering and of course server management is keeping pace with these developments. But that's not all. The architecture and the use of standard building blocks can be extended well beyond the levels that we have talked about. I would like to demonstrate that by actually having somebody who devotes a considerable amount of effort to that to tell us about it. Please welcome Casey Powell, chairman and CEO of Sequent Computer Systems.

ANDY: Hi Casey.

CASEY POWELL: Good morning, Andy.

ANDY: Good morning. Compaq brought us computers. NCR brought us computers and you come empty-handed.

CASEY: Well, Andy. Unlike most of the processors that find their way into computers, the computers that Sequent builds don't quite conveniently fit under the seat in front of you on the airplane. Our computers are substantially larger.

ANDY: Don't they have a bigger airplane than that.

CASEY: Our computers are substantially larger. We have one in a video here. Perhaps if we roll the video you can get a feel for how big they are. Sequent was formed fourteen years ago. You can see here that it doesn't always rain in Oregon. We have been hard at work pioneering symmetric multi-processing architecture that really leverages the price performance of the Intel microprocessor and pioneered that over the years. We have an installed base of over 8,500 systems. And those systems have found their way into many data centers around the world. Recently we introduced the NumaQ* 2000. It's an architecture that really capitalizes on the new SHV architecture from Intel and allows us to put up to 63 quads together in a large single system image so that the user doesn't really have to understand or know anything about the fact that there's more than one quad in the system. We've been very successful working closely with Intel running obviously a very important benchmark which is the SAP benchmark. We've recently completed that benchmark and running it now on only eight quads. You know, far from anywhere near the top end capacity of the system, we've been able to achieve 2804 standard SAP users which we think is a real milestone.

ANDY: Explain to me and put it in context what 2804 users mean.

CASEY: 2804 standard SAP users, just as a reference point, 3400 users was the previous record and that was done using four IBM 390 mainframes.

ANDY: --so your 2804 processor based system is getting very close to--

CASEY: Right. And it's still as I mentioned with only eight quads. We had to rush to get the benchmark done since the architecture is relatively new. We look forward to bigger and bigger benchmarks. At this point I will commit to you that we'll have the high water mark eventually.

ANDY: Well, that's very reassuring, Casey, because Intel is relying on your product as the backend server and your scaleability translates into satisfaction by us as a customer of computers based on old products. So we're counting on you.

CASEY: Well, Andy, we're not going to let you down. We think in order to be successful in the space that Sequent really occupies today which is--we believe the best alternative to the mainframe in the marketplace today that we need essentially two elements to do it. One, we absolutely have to have an Intel processor and we think that anybody who will be successful in that space must just because of the economics. Now we think, even though it's never been confirmed by Intel, that you guys have about 94% market share and have shipped somewhere in the last year--

ANDY: Look at this poker face.

CASEY: 87 million chips. The next closest contender would be Power PC with somewhere around 3 1/2 million. And knowing what you're spending these days, to build wafer-fab facilities, we're happy to ride your coattail on that one. So we think that the game is over for microprocessors. We believe that that's the only way we can go economically to do the job. Secondly, we think in order to be successful in the data center that any company who would be an alternative in the data center today has to have a path to NT in the data center. And we will have a machine in the Deschutes time frame that will allow---

ANDY: Deschutes?

CASEY: Deschutes, yes.

ANDY: What is Deschutes?

CASEY: Deschutes, even though you don't acknowledge that either, is the next processor on the way to Merced. Okay. It'll be the next big Intel processor release. And I love doing this, getting Andy up to speed is always fun. — In the Deschutes timeframe we will have a machine that will concurrently run both distributed shared memory, Unix* and distributed message passing NT on the same machine. In the Merced timeframe, you'll be able to reconfigure that machine to run NT or Unix on different quads from the console. We think this is the way people will want to go, through Unix to NT in the data center. We think it's the safest migration path.

ANDY: Okay, Casey. I look forward to that and we intend to supply you with the best chips so you can supply with a backend system that keeps up with that.

CASEY: Thank you. And we're proud to be a supplier of SAP systems to Intel. Thank you, Andy.

ANDY: Thanks again.

[Applause]

ANDY: The second mega-trend of computing that is very important to us is that all computing in the future will be network computing. And there are implications of that of course in the data center and IT applications. Lower total cost of ownership requires that all the PC's and servers on the network should be managed and managed centrally. The way we think this is going to unfold is there is going to be a spectrum of managed PC's ranging from very simple network computers to basically full function PC's but with new user loadable software called Net PC's to managed PC's. All having the same management models and extending all the way to standard high volume servers again in the same management environment. We think this model is going to be the prevalent way computing is going to be displayed. Networks. And depending on user and software--user requirements and software requirements--there's going to be a whole spectrum of microprocessor powers at each of these points on the horizontal axis. This is a focus area for our development both from the standpoint of platform design and from the standpoint of management architecture. We're very excited on the impact it is going to have in corporate IT infrastructure. This is the internal implication of networking.

The external implication of networking is so obvious at this point that it has become a cliche and we forget how fast it has come upon us is that the Internet delivers a universal network for connected PC's. The growth of the Internet is phenomenal even by PC standards. This is not even the latest set of numbers. The latest set of numbers imply, that in 1997, just in the US, we have exceeded or we are exceeding 80 million users connected to the internet. And consequently, this type of ubiquity implies that a strategic inflection point in terms of business to consumer and business to business transaction. There're two fundamental business process driving forces that make this so. One is that the Internet enables a small business to act like a large business. A year or so ago there was a little cartoon in the New Yorker that says "On the Internet, nobody knows you're a dog." A more polite modification of that "On the Internet, nobody knows that you're a small company." You act, look and feel and can do business like a large company and this is of immense importance in what is the most importance factor in the success of any business, which is customer acquisition. At the same time, as small businesses begin to look like big businesses and become a competitive force in the life of big businesses, the Internet provides the defense mechanism for large businesses through the means of internal collaboration: they give an opportunity to fight back and overcome the disadvantage of a large business which is a slower speed of operation.

I would like to end by giving you a demonstration of the first item of how small businesses can look like large businesses and provide a seamless business process operation of customer acquisition, customer satisfaction, using a technology that is the foundation of a product of a joint company that was formed about two weeks ago between Intel and SAP called Pandesic. Pandesic's purpose is to exploit the Internet toward the adaptation and modification of business processes that allow small companies to act like large ones and provide a complete electronic business solution that includes hardware, software, installation support for merchants who want to operate on the Internet. I would like to ask Julie Sackman from Pandesic to join us here and give us a demonstration of what the Pandesic solution is about. Julie. Hi.

JULIE: Hi Andy. We're very happy to be here. Thank you for letting us show you Pandesic. Before we get started, what I want to tell you a little bit about is the actual demonstration that we're going to go through a little bit about the business process. MountainLake Outfitters, a fictitious business, purchased and installed a Pandesic product. They're running an Internet store and they sell camping supplies. Camper Co. is a partner of MountainLake. They have committed to sell products through MountainLake's Outfitter Warehouse. What we want to do here today is take you through the entire demonstration. But really an important part of this is that while these business partners are located here physically on stage, it's important to point out that they can be located anywhere within the United States. So why don't we go ahead and get started, over here at the consumer end, and I'd like to invite you to be the consumer today.

ANDY: Okay.

JULIE: Great. Do you enjoy camping?

ANDY: I've always wanted to resume my camping career. Who needs hotels?

JULIE: Well, one of the things I want to show you here a little bit before we get started is that when we go through this business process, a couple of things I'd like to point out to you is that through the power of integration here and through standard browsers, we are able to run this business for MountainLake as well as their business partners seamlessly. So why don't you go ahead and log on to our homepage. We've got you physically located in Folsom, California today.

ANDY: So this is, as it were, my home computer. I'm sitting home. I've found MountainLake Outfitters. And I want to buy something from them.

JULIE: Yes, that's correct.

ANDY: So I get on the MountainLake Outfitters page.

JULIE: Yes. And what we've done is we've taken the liberty of setting you up already as an existing customer. So why don't you go ahead and shop as a member only. And what we've done is we've set up your member number to be 84.

[Tape Side B]

JULIE: At MountainLake's Warehouse. You bet.

ANDY: Mountainlake Warehouse.

JULIE: And that's located today in Sacramento, California. So when you hit that, within seconds and the ability for us to be able to do that, we don't think anybody has that capability. That offers a merchant to be able to put their products on the dock within thirty minutes.

ANDY: Well, the only way we could demonstrate that is other than you just saying it is if we could hear the printer whir up when I submit the order.

JULIE: Amazing things will happen. Why don't you submit that order?

[Computer sounds]

ANDY: No kidding.

JULIE: And there we are at the warehouse.

M: Here we are in the Mountainlake's warehouse. And as Julie said that picklist did print out in seconds. We can go ahead and take a look at some of the information on here. It's got Andy's home shipping address and we've got the product information including the Bin location of that product where we can find it in the warehouse. So let's just take a second here and log into Mountainlake's Pandesic system. We can get there just using a standard web browser. So I can go ahead and log in. And once we're logged in, we can see we have a couple of choices here. We have goods received and process delivery. And since we're going to process Andy's order, we'll go ahead and click process delivery. I've got a web scanner here that I can use to just go ahead and scan the bar code. When you hit enter, we can see it brings back all the information that is on the picklist to confirm that everything is correct. Our carrier has provided us with some preprinted airbill numbers and we can go ahead and scan that in. It'll help us track the delivery. And Andy can come back later and log in and check his order status if he wanted to track that shipment. Now as his delivery is processing the picklist is going to come up on the printer, the packing slip will come up on the printer. And it'll give me a moment to here to go ahead and pack Andy's box up.

ANDY: Do they use packing material in this warehouse?

M: We take a look at the packing slip here. We can see it's got some preprinted labels on here that we can stick on the box. We'll go ahead and peel one off right now. That's also got a return label in case Andy needs to return that product and it's got some information we ask him to fill out if he needs to go ahead and do the return. So we'll go ahead and pack this up. Place it in the box. Put our airbill number on here for the carrier. Pack up this box and it'll be ready for our next carrier pickup and it looks like we've made it just in time.

[Laughter and applause]

JULIE: As you saw, what we just showed you here was really the front-end process but what it demonstrated pretty effectively is the level of integration that you saw. Standard consumer experience pretty much on the web but the certainty, you received a delivery date and you actually were able to get that shipment started in the warehouse quickly. We demonstrated that piece for you but let's take a walk over here. I mentioned to you that Camper Co. earlier has committed to sell products through Mountainlake Outfitters. Let's take a look at what happened to their level of supply when that order was placed. So over here to show us that is Serena with Camper Co.

SERENA: Hello. Thanks, Julie. As a result of the order that you just placed, I received an email notification telling me that canteens have fallen below the minimum inventory.

ANDY: I just have to state that you are the supplier to Mountain---

SERENA: I am the supplier to Mountainlake. I am Camper Co. And I have products that I have committed that's located at their warehouse. Okay? Because canteens has fallen below the minimum inventory, Mountainlake's Pandesic* system has sent me an email notification. Here is the reorder notification. It tells me that right now I have currently nineteen units in stock. Now what I want to do is I want to check my inventory status so I'm going to log in to the Mountain Lake's web page for the supplier and that, I will have my own log in and all I need is a standard web browser to do this. Because I'm interested in canteens, it's underneath camping supplies. I'm going to find canteens and now I can see real-time the actual inventory that's in the warehouse right now. And it says nineteen. Now I notice that canteens are selling at a faster rate so I want to increase my minimum inventory level.

ANDY: So customers like me that buy three at a time---

SERENA: That's right. It's the height of the summer. It's hot out there so--Now I want to increase it to thirty. Now when I make this change, the next time canteens fall below thirty, I will get a notification again from Mountainlake's Pandesic system.

JULIE: What else would you do, Serena, based on this information?

SERENA: Well, based on this information I now know I have to restock the warehouse. So at this point I would go into my currently legacy system and log in. And I want to send--I need to replenish consignment. My consignment inventory over at the warehouse. And it's going to be Mountainlake's location. I want to send a case over which is 24 and I want to send--So by submitting this request using my legacy system, it's going to initiate a goods movement and send more canteens over to the warehouse.

JULIE: Now, Andy, what I want to point out is that we came all the way from the consumer and the warehouse relationship. We kind of saw that through Mountainlake. But now this demonstrated very effectively the real time capability, the real time exchange of information which allows Camper Co., one of Mountainlake's business partners, to run their business and actually change their business conditions as a result of the information. They have the ability to react quickly and pretty seamlessly. Now, we've talked about this but what's happened with Mountainlake? What's going on here with the person or the company that actually purchased a Pandesic solution? Why don't we go over here and talk to Brian and take a look.

BRIAN: Great. Welcome to Mountainlake, Andy. What I'd like to do is show you my Pandesic system. Not only does it include the software that we've been able to demonstrate to you here today, but it also comes packaged with the hardware. What we have here today is actually an HP Netserver* LX Pro. And it's configured with four Intel Pentium Pro processors running at 200 megahertz. Now what I'd like to actually do is run the Pandesic billings process for you. Now keep in mind, I'm going to do this for you manually but in a standard implementation this would certainly happen automatically within the system. So when I go ahead and start this transaction, what you're looking at is a list of all the different shipments that have gone out of Mountainlake warehouses today. And here's your order right here. Now when I complete this transaction there's a couple different things that are going on in the background. The first one was the completion of the credit card processing. When you actually placed your order there at your own workstation, we went out through the cybercash gateway through credit card and actually got an authorization number. At this point we're going to use that authorization number and go out through the cybercash gateway again and actually charge your credit card at this point. The second thing we're going to do is actually generate an electronic invoice that will be sent to you via email and it basically acts as a confirmation that your product has actually shipped out of the warehouse and that your credit card's been charged. Now we've heard us talk a lot about Pandesic integration. A really good demonstration of that is an accounting view of the billing document that we just created. And if we take a look at this, what you're looking at is a list of all the different GL entries that have taken place based on the business scenario that we just took you through.

JULIE: What did we make on this order?

BRIAN: Yeah, that's a great question. What I'm going to do is flip over to a standard web browser and I'm going to run one of Pandesic's merchant reports for you. And when I go ahead and bring up this report, what you'll be able to see is the Mountainlake balance sheet profit and loss statement. And let's go ahead and page on down to the bottom and we'll be able to see that we've made just about $15 on the order that Andy placed. Again, real time integrated information that allows me to run me e-business solution.

JULIE: That's super. Pretty terrific. So actually now that is our business scenario today. And what you just saw was an example as Brian said. The level of integration, the real time flow of information allowing the business partners to act and react quickly and also the dynamic financial reporting which is the key part of Brian's business as he's running it from Mountainlake. So we'd like to thank you.

ANDY: All of this is wonderful but where's my canteen?

JULIE: Should be coming. Federal Express Overnight.

ANDY: (Delivery person delivers order on stage to Grove) That's wonderful. Thank you very much.

ANDY: Thank you. Hey guys. Brian. Would you take it back from me? Thank you. The key thing that I would like to stress here is only the merchant needed to buy a Pandesic system. Everybody else has a web-connected computer. Any web-connected computer can access the Pandesic system from home, from the warehouse or from anywhere else. And what we have in the world today is, over the world, 250 to 300 million connected computers capable of accessing the web. We project in a few years time, we're going to be talking of the order of a billion web-connected computers world wide. That is going to be the medium in which a merchant by the purchase of an e-business solution like Pandesic will be able to use to their satisfaction to act like a large company. This demonstration is really intended to bring to you the glimpse of the biggest mega-trend of them all. In the future, with a medium of hundreds of millions pushing onto a billion connected computers available for commerce as well as for other tasks. Transactions, business will be connected not face to face but stream to stream. And I think this will be the mother of all strategic inflection points because it's going to change the way those of us in the information technology industry do our business but even more importantly, it is going to change the way everybody in commerce does their daily business. As with strategic inflection points there are two paths. There's a path of ascent and a path to descent. And which path you are going to be ending up on depends on the decisions and the plans and the implementation of those plans that begin at the time of that inflection point which I submit to you is today.

Thank you very much for your attention....

* Other names and brands may be claimed as the property of others.