The eXCHANGE

Paul Otellini
San Francisco, Calif., USA
October 11, 2000

PAUL OTELLINI: Well, good morning. I think what Craig described is really all about the promise and the potential of this event. Quite simply, it's about the industry coming together to solve the very real problems that all of us have in building out our e-Businesses.

If you think about the industry, it really has changed quite a bit in the last 20 years. In 1980, the computer industry was roughly about $80 billion. It was comprised mostly of companies that produced proprietary stacks: hardware, software and applications. There were fairly few vendors, and each of them focused on delivering solutions, but solutions which really didn't change your business processes, but, rather, allowed you to automate what you had and scale your business upwards.

This is what Andy Grove described a number of years ago as the old computer industry. Well, the PC changed all that. The PC enabled a new computer industry. And, in fact, that new computer industry has grown by a factor of ten in the last 20 years, to over $800 billion a year today.

As Craig described, it grew that way, and it grew that large, because it embraced an entire industry, the creativity of engineers in many companies, developers in many companies around the world, all of them innovating on standards, all of them delivering solutions that were interoperable, all of them trying to work together. What the PC enabled, fundamentally, was personal productivity. It empowered the individual employee to do his or her job better.

But the PC can't live in a vacuum. What has to happen is the PC evolves in its capabilities. They became connected to each other in departments on LANs, they became connected to other employees in companies around the world on wide area networks.

This drove a requirement for a new class of servers. Standard, high-volume servers arose. And they did the fundamental work associated with running this class of the enterprise, the file servers, the print servers, and the mail servers that made our mission critical applications start to work.

But the back end of the companies didn't change. It really was very much the same as it was circa 1980, large mainframes or mainframe class machines running the legacy applications, the high transaction applications required for the billings processes or the data tracking required to run your business.

And the real time needs of the people on the ends of the LAN and the WAN sometimes collided with the batch requirements associated with the legacy systems. They became increasingly difficult to deal with your job and to get the data you needed day to day.

What I believe we are witnessing today is the birth of a "new" new computer industry, one that is as fundamentally changed from the last one we saw as the old one prior to that. And, in fact, we believe the potential of this "new" new computer industry is as large, relatively, as the last one. In fact, we believe it has the potential to grow to in excess of a trillion dollars, another factor of ten growth over the next 20 years, comprised of products and services by people who participate in the computer industry.

This "new" new computer industry is different than the last one. Instead of focusing on individual productivity, what it focuses on is enterprise productivity. And at the end of the day, instead of empowering just the employee, it empowers the customer. Because at the end of the day, that's what drives everyone's business.

But to take advantage of this change and to be able to scale your business from where you are today, to be able to grow to become the "new" new computer company participant, you need to be able to do a number of things. You have to scale out your infrastructure. And, more importantly, you have to change your business processes to be able to become much more customer centric. If you don't become much more customer centric in your processes, your competition clearly will. And at the end of the day, your customers will demand it.

What I'd like to do is take this from the abstract, though, and bring it into reality and introduce you to someone who represents a company born on the Web. Let me bring out Bob Linsky, who's the vice president of MIS and operations for DoubleClick. Good morning, Bob. We've been talking about customer centricity and scale. And I think are you a poster child for both of those. Can you tell us a little bit about DoubleClick.

BOB LINSKY: Sure.

We are a company born on the Web. We are a global Internet advertising company servicing both the advertising and publishing communities. Basically, what we do is provide media and technology expertise and make advertising work on the Web.

We were founded in 1996 by Kevin O'Connor and Dwight Merriman in the basement of Kevin's house. We built the technology on one NT platform. And we scaled that platform today to serving over 1,500 servers worldwide.

PAUL OTELLINI: How did you make that scalable result happen?

BOB LINSKY: It's great. Because I actually do run the operation. Not only do I get to talk about it, but I walk the walk as well. It's been a lot of fun. What we've done is built the applications so we can scale the architecture both horizontally, which means we add servers against the applications, as we need to scale. And we also build the applications vertically.

What we've been able to do, basically, is continue to add boxes and add servers and CPUs to make that application run. So, as I said, 1,500 servers, ten data centers domestically, 13 internationally. And I'm going to guess somewhere around 6,000 CPUs running in that environment.

PAUL OTELLINI: Did you have any trouble building out that kind of scale?

BOB LINSKY: You know, the first 50 million ads a day wasn't a big deal. And that was around 1998. Just to give you perspective, we're doing about 2.5 billion ads a day today. And we'll run 44,000 transactions a second coming into our system. It's continually being a challenge to scale the systems. And with that, we look for architectures that have open standards, that the vendors have the ability to move applications onto easily, and that we can look for price/performance curve characteristics to continue to meet our needs.

PAUL OTELLINI: Going forward, what kind of technologies are you looking forward to seeing from companies like Intel?

BOB LINSKY: You know, it's an interesting question, Paul. We hit a plateau at one point. We thought we had burned out NT 4.0 and the Intel platform. And we had concerns. And we worked with your Intel Services organization, the Solutions Center, and we learned about your technologies. We learned about the 36-bit addressing technology, so that it gave us more memory capability and addressing space. We learned how to take advantage of Windows 2000 out of the NT 4.0 environment. We've seen tremendous benefits. What we're really looking forward to is the IA-64 and the Itanium™ environment and running that into our back end systems.

PAUL OTELLINI: What will that give you?

BOB LINSKY: What we believe it will give us is the ultimate price/performance, industry standard, and competitive model that we're looking for to run not only the front end pieces of our systems, but also to finally move some of those what we've now created as legacy systems, or those monolithic systems into that environment and do that efficiently.

PAUL OTELLINI: Got it. Very good. Good luck scaling up here.

BOB LINSKY: Thank you.

PAUL OTELLINI: What Bob described was really how a company that's born on the Web builds out an infrastructure. Not all of us had the luck of being born on the Web. Many of us have to deal with the infrastructures we were dealt, that is, the legacy systems or the PCs or the LANs and the WANs that are out there today.

What we envision is that as companies move to embrace e-Business, there is a relatively common pattern of deployment of infrastructure that we're seeing. And that common pattern is really represented in the slide that's up here now. It is moving from what was described historically as a two-tier model to a three-tier model. And that's not just because the industry wants to sell you more servers. It's because the structure actually does significant benefits for you in deploying e-Business.

The first thing that changed here is really the Internet. The Internet allows or requires companies to not just deal with static clients that happen to be out there in terms of PCs. Customers can come from anywhere on earth. They can come 24 hours a day, seven days a week. And they can come from many, many devices. They can come from cell phones; they can come from PDAs, or, of course, from personal computers.

We have to be able to build an infrastructure that can take inputs from all of those customer devices into our business to allow us to take the orders and service the customers.

That drives a change in terms of the infrastructure. The infrastructure that we talked about before in terms of these relatively small, standard, high-volume servers that did things like file and print and mail, really has evolved. And it's evolved to handle applications, like you saw with CyBerCorp earlier, that are essentially the front door of your company. They allow you to configure the operations of your company. They do tasks that didn't exist a couple of years ago, things like security, things like directory services, caching, load balancing, all the things you need to make sure that that front door is open and responsive to your customers 24 hours a day.

But a lot of the energy of the industry that's happening right now is focused on this middle tier. Because the middle tier is really the personalization of e-Business creating the character of your company.

What's happening here is that this is where the applications all reside. This is where you can actually conduct the business that you want to do with any given customer online. What we're seeing here is whole new classes of applications being developed around new classes of hardware. That's really what the bulk of the demonstrations that you'll see in the showcase in the next hour are all about.

We really haven't forgotten about the back end. And, in fact, those databases still exist. Many of them are larger. Many of them are even more important now in terms of determining characteristics about our customers' buying patterns or about their needs going forward.

And, of course, all the transactions still lie there, whether you're a bank or whether you're a company trying to do its billings. Many of those reside, the storage, back in the back tier.

What we see now is that as new companies like DoubleClick come online, though, they can begin anew, they can begin afresh and build out this infrastructure with new applications and new capabilities. Things like e-markets, sort of pioneered by Commerce One, were built from the get-go on Intel® Architecture.

Our work in the low end, standard high-volume servers, has really paid off well. It's given you, as users, a number of choices. It's dramatically lowered the cost of front-end computing. And, in fact, Intel® Architecture servers now represent over 75 percent of all of the Internet infrastructure servers that are deployed out there today.

But that's not enough. Going forward, we have a vision of moving the architecture and moving the benefits of high-volume economics up the computing food chain. And this is really the promise that we tried to deliver upon when we first envisioned Itanium™ some five years ago. Itanium™ began as a collaborative project with Hewlett-Packard jointly defining the architecture for our next-generation microprocessors.

And when we defined this architecture, we really had two fundamental premises or goals in mind. The first is that as systems deployed, they had to be deployable into the existing infrastructure, that is, that we believe that companies and IT departments were not going to throw out their infrastructure just because Intel and its partners brought out new hardware.

Specifically, we wanted to make sure that this new hardware would take advantage of not only a number of the predominant legacy operating environments that are out there, but also a number of the newer ones that were being developed. So we designed this product to ensure that it had, from the start, support in multiple operating system environments.

The second design premise was to ensure that the solutions that were going to be developed didn't require you to throw out everything you had built in the past.

We know that 75 percent of the infrastructure is running on Intel 32_bit servers today. There's a lot of proprietary and non_proprietary applications running on those servers. We wanted to make sure that those applications seamlessly melded into an environment that also included Itanium™ processors for specific performance over time.

We also architected it for the future. We looked out over time when we designed this product line and decided what were going to be the requirements for e-Business for secure transactions going forward. And we built very, very strong capabilities for SSL security and for addressing very large memory databases as systems got deployed, recognizing the need for both security and data_mining over time.

What you'll see in the showcase in half an hour or so is that there are a number of companies already building pilot systems based upon Itanium™, everywhere from single-processor workstations to a 32-processor mainframe class machine from some of the companies that are in there.

The solution stacks are already being developed. There are solution stacks in finance, in manufacturing, in e-retailing, and in entertainment.

And, in fact, over half of the demonstrations in the showcase are running or using Itanium™ processor systems somewhere in their solution stacks.

I think you can tell that I'm relatively excited about Itanium™. I think, though, that you understand that it's not just Intel that's excited about this. End customers who have seen it and seen the potential of these systems are also excited.

I'd like to run a short video from a couple of our potential customers.

VIDEO: CERN is a research organization for physics located in Geneva. And our task is to do all the infrastructure for all this line of research so that all the scientists from around the world can benefit from it.

The Itanium™ processors that we're using today can help us build what we see as becoming the power plant of the future information grid. Today, with our choice of Intel-based computers, we get, in a single node, as much power as not that long ago we got in a supercomputer. But even better, we can actually replicate the systems and put hundreds of them to work on the problems all at the same time, using these architectures.

Bid4real is an online real estate marketplace that enables sellers to auction their properties direct to the buyers with or without a broker. A typical Bid4real seller is one that recognizes the power of the Internet to accelerate the transaction cycle. That, in turn, saves them money.

Bid4real had a tall order. They wanted to build the world's best auction site online.

We had to ensure that both buyers and sellers had a comfort level in the security of the transactions being conducted. These assets are very expensive. The Itanium™ processor allows RSA SSL to run secure much faster, five to seven, sometimes ten times faster than processes.

Wells Fargo has about 6,500 ATMs. And we process roughly 380 million ATM transactions a year. We look at where customers actually do their banking, at the teller ATMs, online banking and phone bank. Gives us a tremendous amount of information about our customers if we can distill it into something that's manageable. The Intel® Itanium™ processor will allow us to understand our customers better. Our goal is to match this Itanium™ processor with SAS software, which will give us a winning combination, because it will allow us to analyze a lot of data very quickly and understand our customers better. After we test the Itanium™ processor this year, we plan to deploy it across the bank in 2001.

PAUL OTELLINI: Now, we're particularly excited about some of these specific applications, you know, things like the CERN activities that take advantage of the supercomputing capabilities on this chip. But also things like Wells Fargo, which allow us to bring our architecture into application spaces where it really has never been before.

And to give you a little bit better idea of how this is going to happen at Wells Fargo, I'd like to bring out Robert Cross, who's the executive vice president of technology for SAS.

Good morning, Robert.

ROBERT CROSS: Good morning.

PAUL OTELLINI: How are you today?

ROBERT CROSS: Very well. We appreciate the opportunity to share with you some of the solutions that we're working on and the work that we're doing jointly with Intel.

Our solutions scale to all of the processor platforms that we operate on and our collaboration with Intel has helped us do that on your evolving line of processor families.

PAUL OTELLINI: That's great. Maybe you can show us how it works.

ROBERT CROSS: Absolutely. If we can step back behind the monitors, I'd like to introduce Gary Mailer, who will be helping us with this demo.

On the screen in front of you, Paul, you see a Web site. It's a personalized Web site for an online banking system. A customer has logged on and has been offered a new car loan. That car loan isn't offered to everyone. It's offered to her individually, because that Web site understands that she is a good prospect to buy a new car, and she's already approved for a new car loan for that amount.

This is the kind of personalized interaction that we'd all like to offer to our customers and that we'd like to use ourselves.

PAUL OTELLINI: I agree. How do you go about building this kind of personalization?

ROBERT CROSS: That's the key question.

How do you do it, not just in the small, but in the large for a customer like Wells Fargo that has millions of transactions and years of historical information about their customers that they want to bring to bear?

In our case, we'd like to take a look at some of the analytical tools that we offer. What we'd like to show here are the tools that we offer for business analysts to understand their customers, to be able to wisely offer the products and services in a personalized fashion.

Here, we're taking a look on the screen in front of me at our data mining tool. It's the tool that you would use after collecting and organizing and summarizing your data warehouse in order to understand the huge amount of data, the gigabytes or terabytes of data that you've collected about your customer transactions, Web transactions, things of that kind.

And this would allow you to understand what's important and to make decisions about what to offer your customers in a personalized fashion.

We have a simple credit risk analysis here. We have trained it. We have subsetted the data, we have trained that neural network to properly select good credit risk candidates. We've validated it against the entire database. And we can see here on the plot in front of us that this simple model that we started with is accurate about two_thirds of the time.

PAUL OTELLINI: I would think that two-thirds of the time, the bank's not going to make a lot of money on this. How do you improve this rate over time?

ROBERT CROSS: That's correct. And the next tool that we would use is an exploratory visualization tool called SAS Insight. We started with a very simple model that we could improve. What we see here is, with this multi-dimensional navigation, visualization, and analysis tool, that we can take a look at the data. We see that the initial graph here doesn't really tell us very much. But if we use the tool to rotate the data, we can see, as we move it around, that the vast majority of the data, two-thirds of it, does in fact line up along the response surface that was predicted by the model. It's a very simple model that just involves the age of the candidate, the amount of the loan, and the amount of time they've been in their current job.

And so we can use this tool to take a look at the outliers, for example, and understand if those are irrelevant customers somehow, or if they represent customers that we need to include in our model. And we can iterate between this tool and the data mining tool to improve the predictive quality of our model.

PAUL OTELLINI: So an analyst could use this real time to be able to more finely hone that neural network in terms of improving the loan rate?

ROBERT CROSS: That's correct.

PAUL OTELLINI: That's great. What else does this technology do for you?

ROBERT CROSS: Well, what we have seen is the advancements in Intel® processors, such as the Pentium® 4 processor, have allowed us to dramatically improve the performance of these tools.

This tool in particular is capable of partitioning in its work between the exploratory portions, just as the data mining tool, you can perform the interactive processes on this Pentium® 4 processor workstation, and you can put the heavy lifting of the large amounts of data and the varied floating point and data-intensive functions on the data server behind us.

We've been able to use the new floating point instructions, streaming SIMD and cache instructions on the Pentium® 4 processor to dramatically accelerate these tools.

PAUL OTELLINI: I can see how this takes advantage of this high-performance client, but what about the back office? How does that take advantage of Itanium™?

ROBERT CROSS: That's key. This entire analytical process is predicated on the power of the server. You have to collect, summarize, and then analyze the data warehouse. You may have to extract focused data marks.

And, again, the heavy lifting is being performed on the server side of this connection.

This is where the Itanium™ really comes to the forefront.

In the screen in front of you, we can see some performance tests that we've been doing. We have results on a Pentium® III Xeon™ processor. And as we perform some of these tests, we can see the results for Itanium™. PAUL OTELLINI: Your benchmark is your existing Pentium® III Xeon™ based systems against the new Itanium?

ROBERT CROSS: That's correct.

And as we see these results come up on the screen, we can see that the Itanium workloads run about one and a half to four and a half times faster than on the Pentium® III Xeon™ system.

And we have to bear in mind that these are early Itanium systems being compared to a SAS system that is highly optimized for this Pentium® III Xeon™ system that we use as a baseline.

These test workloads are representative of database creation processes. They're very I/O intensive, some of them are very computationally intensive because of the more complex queries and variable computations and so forth that are involved here.

PAUL OTELLINI: So as you continue to tune and optimize the compilers, you'd expect the blue lines to get even shorter here?

ROBERT CROSS: That's absolutely correct.

PAUL OTELLINI: What else does this technology give you? What are the real benefits for you?

ROBERT CROSS: The two key takeaways that I come from, from these results, are, one, that the Itanium explicitly problem design concept actually works.

PAUL OTELLINI: Good.

ROBERT CROSS: -- that the design making the silicon and the advanced compiler technology, which was very leading edge, coupled to each other to provide this performance benefit is actually a successful approach.

And several of these that are among the most beneficial, the fastest on Itanium, have been compiled with the profile-guided optimization techniques on the latest Itanium compilers. And they clearly show tremendous benefits. All of these workloads show tremendous benefits from the large memory and from the advanced floating point capability of the Itanium.

PAUL OTELLINI: As you go to deploy systems like this in places like Wells Fargo, what kinds of degrees of flexibility are you looking for?

ROBERT CROSS: Well, I think the key thing is that the SAS system on Itanium scales to the enterprise. And it's going to make it possible for customers such as Wells Fargo not just to engage in wishful thinking with regard to their personalization, but to actually execute on their vision. And together, the SAS analytical tools on Itanium are going to make it possible for them to execute, to realize their goals, and to deploy those features to customers in the next year as these systems become generally available.

PAUL OTELLINI: Now, one of the intriguing things about the SAS environment is, as I understand it, you've ported it to run on multiple operating systems, so, again, back to the point I made earlier about living with the legacy installations people have, this seems to fit this model.

ROBERT CROSS: Absolutely. The SAS system is portable to a wide variety of platforms and we will be supporting AIX 5L and HP-UX as well as Windows 2000, which we've used for our demonstration today. And it is very possible to connect to the legacy data on those other older operating systems and to old TP databases and collect the data from the Web sources and analyze it all and provide this kind of personalized service.

PAUL OTELLINI: Wonderful. Thanks very much, Robert.

ROBERT CROSS: Thank you very much.

PAUL OTELLINI: And Gary, thank you.

(Applause.)

PAUL OTELLINI: If you'd like to learn more about this exciting application, there is a SAS booth in the showcase with the same demonstration, and people there to help walk you through it.

What I'd like to summarize is to sort of conclude that leaving you with the thought that Intel really is betting its future on e-Business.

Craig described how we are moving as quickly as possible ourselves into becoming an e-company. But that's only part of the story. It's an important part, but only part of the story.

The real focus of the company is on delivering and building and developing a number of building block solutions to enable this e-Business buildout we've been describing today.

And as we go about doing this and working with our partners in the industry, we're really keeping three design premises foremost in our mind. The first is that we believe in our heart of hearts that customers demand choices. They demand the ability to supply from a variety of sources and to deploy a variety of different types of environments.

We need to stay focused as an industry on enabling solutions. Craig talked about customers not caring about the speeds and feeds of their hardware, but, rather, does it do the job.

And, thirdly, we need to be focused on innovation. And that really is what this event over the next day and a half is about. It's not a trade show. No one's going to be trying to sell you anything, at least not overtly, at least I hope not. And it's comprised much more in a seminar fashion. There's a variety of venues that you'll find, including a number of speakerships. In fact, we have over 50 speakerships at this event, given by over 75 different individuals from 60 different companies.

There are four other keynotes, from IBM, Hewlett-Packard, Microsoft, and Andy Grove wrapping up in a panel discussion tomorrow afternoon.

But the real power of the event is in the showcase next door. And it's comprised of 56 demos. Each demo isn't just an isolated demo, but, in fact, an entire solution stack integrating a variety of applications on many different types of hardware.

There are over 77 demo stations, representing 80 hardware vendors and software providers that are out there, over 220 applications, most of which working seamlessly together, running on four major operating systems.

And it's not just demonstrations. What you're going to see is that these solutions actually represent real, live business environments for companies like Charles Schwab, NCSA, CERN, and Loreal.

And we've tried to focus and bring you environments in many different areas of business, from ERP to CRM to supply chain management all the way down to entertainment at the end of the day.

So on behalf of Intel, I would like to welcome you to The eXCHANGE, to welcome you to the birth of the "new" new computer industry, to go out and experience for yourself the kinds of solutions that are being built with customer interoperability in mind, software that is being built with the customer interface in mind, networks that are converging in their many data types to common protocols and common interoperability. And, from our perspective, hardware and system solutions that are modularizing and standardizing.

Because in that modularization and standardization are two things for you as a customer. The first is the sheer potential to be able to scale your businesses to grow as fast as you can. And the second is the thing that has always been brought to you by the horizontal industry, and that is the volume economics of scale. The sheer size of which we all operate on gives you, as buyers, the best price/performance and the best range of solutions.

With that, I'd like to turn the audience loose.

The showcase is out the front door of this pier, to your left about 200 feet. And it's in an joint and problem pier to this one.

You have about an hour and a half to take the first experience of the showcase. There's a number of events happening simultaneously in there. And the next keynote is back in this auditorium in one and a half hours. And it's Carly Fiorina from Hewlett-Packard.

Thank you very much for coming.

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