 |
 |
 |
LinuxWorld Conference & Expo Sean Maloney CHARLIE GRECO: Thank you. I want to take this opportunity to welcome everyone to LinuxWorld Conference & Expo. We're very excited to have you here this week, and we have a tremendous amount of exciting activity going on this week, and we hope you enjoy your stay here in San Jose. Beginning with this keynote, I'd like to introduce Sean Maloney who is a senior vice president of Intel Corporation. He has been with Intel since 1982. Sean began his Intel career in Intel European headquarters where he spent nine years, first as Intel UK manager of applications engineering, then as country manager of Intel UK, and director of marketing for Intel Europe. Most recently, Sean was general manager of Intel's Asia-Pacific operations. From 1992 to 1995, Sean served as technical assistant to the chairman and chief executive of Intel, Dr. Andrew S. Grove. In 1995, Sean moved to Hong Kong to manage Intel sales and marketing activities in Asia-Pacific. He returned to the United States in April of 1998 to take his current position. He was promoted to senior vice president in January of 1999. It is my distinct pleasure and honor to introduce Sean Maloney. (Applause.) (OPENING VIDEO PLAYS) SEAN MALONEY: Well, good morning everybody. As the SlashDot guys say, farewell, adieu. I say good morning and welcome to LinuxWorld. Actually, we liked the idea of closing of the video with that clip because it was such a funny clip it was difficult to give a stiff, formal technology presentation, and what we wanted to do today was have a wee bit of fun. We actually laid out the stage, it's a bit like a game show and behind each one of these doors we've got technology demonstrations. And we've got some pretty exciting stuff here. I want to start by saying a big thank you to Larry Augustine and the VA guys who have given us a lot of help over the last few weeks. Many of the companies you saw in the opening video – all of the applications are accessed across the Net. Some of those companies have actually started or developed to take advantage of the Internet economy. I'm going to spend most of the time here looking at technology demonstrations, on how we address the challenges of the Internet, and how we go about making the computer systems more robust and reliable. But I want to start by doing a little tour back over the microprocessor. On a march or a path towards what we call "free" MIPS, and that's free in quotes The idea came from back in the early 1990s when we were looking at the avalanches of processing power that were building up wave, year after year, and Andy Grove who was then our CEO came up with the idea that there was a fundamental relationship between processing power and communications bandwidth. And as more and more MIPS flooded out around the world they would inevitably suck in more and more communications power and you would have a reduction in communications costs. And somewhere these kind of free, very low cost MIPS, would collide with low cost communications. A few years later now you look back on it and as a computer person you complain that communications costs haven't come down as fast as we would like. But if you look on the processor side, I think there has been pretty considerable progress towards massive, massive, massive reductions. The story obviously started back with the 4004, which was in the early 1970s, the world's first microprocessor. All through the '70s a series of iterations, the 8085, 8086, , 8088. By the time I joined Intel in 1982 we were on the 286. It had some built-in memory management features so they started to get some early ports of Unix, operating systems like Linux running on it. And I joined Intel as a software guy working on Unix on the 286. This is an incredibly damaging start to your career. (Laughter.) SEAN MALONEY: I've forever been left with the delusion late at night that I can get more done on VI and a line prompt than any piece of software written afterwards. But we were kind of hanging out for what seemed like a long period of time, for a good 32-bit engine, Intel's 32-bit engine. That came along with the Intel386tm processor in 1985. And the Intel386 processor, of course, was used by Linus in the early 1990s in his original work on Linux. Subsequently, the Pentium® processor came along in 1994. That was pretty much the year there was the whole multimedia boom. The volumes in the PC industry began to get very, very large. And we're now in a situation where we're exceeding the 100, 120 million units a year. That Pentium processor era was the multimedia era, and what we're seeing now with the Pentium® III processor is just tens, hundreds of millions people coming onto the Internet and primarily using the PC to do it. Gordon Moore who was one of the founders of Intel has been playing with the idea of looking at what's happening with all these transistors, and the PC has sucked along behind it this huge industry producing transistors from memory and so on and so on. And actually, if you look at it, the transistor which is a post-war development within 30 miles of where we're standing here, has gone from extremely low numbers around to a cumulative level of 10 to the 17 transistors shipped. Just to get that in some context, Gordon wanted to give it some context so he decided the one thing there are a lot of in the world is ants. Now, I don't know how he calculated the number of ants, but he's pretty precise on his numbers. Anyway, 10 to the 16 is apparently the number of ants there are. So for every ant, there are now 10 transistors. And that gap is actually increasing because transistors breed faster than ants. Now, another way of looking at it is, of course the way the price has been coming down and the discrete transistor which started off in the multiple of dollars is now somewhere around the micro-cent and the milli-cent. So incredible amount, more and more transistors being squeezed into a smaller, smaller space. More and more processing power. And you end up with a phenomenon that if we were to stop the clock now and look again in 12 to 15 months' time, the amount of processing power we ship will be more than the entire total number of MIPS since the development of the computer at the end of the second world war. And you can make that observation at any point looking forward over the next 12 to 15 months. So in a sense, there has been progress towards "free" MIPS. That first Unix machine I worked on, microprocessor-based Unix machine was about $5,000 and it performed at about a half a MIP. And now if you add up a fully loaded PC with a Pentium III processor at 500 or 550 megahertz, it comes down to about a dollar fifty or $2 per MIP. So there has been huge progress towards "free" MIPS. But of course I do say free, as many of you will point out, not open. (Laughter.) SEAN MALONEY: Okay. But life isn't about what have you done for me so far. It's about what are you going to do for me now. Intel is basically a manufacturing and technology company so we took the camera into our advanced lithography labs and our extreme ultraviolet labs, and went around and asked a few of the engineers there what excites them and what they're working on to keep this thing pushed out through another three generations or so. (VIDEO QUOTES PLAY) SEAN MALONEY: That was a quick tour around some of the work that's going on. But I think we're all going to need a lot of smart, energetic, hard working people. Because in many ways we're sort of at the end of the beginning of the computer industry in that I think now, really, the fun really, really starts with the arrival of the Internet. And the challenges that 200, 500 million or a billion users will bring to us are really, really huge. They dwarf any of the previous challenges that we faced as an industry. And I think if you look at all those challenges together and try to sort them out into a list, I think there are a series of things that we need to do. We need to have access anywhere to the Internet at an affordable price. We need to have a way of getting through the sensory overload, the vast amount of material and e-mails that inevitably is going to grow and grow and grow. We need some way of handling that. We need to get instant response and total reliability. And I think those are really the list of things we need. We never want it to fail, we need instantaneous response, we need the information we want exactly whether he we want it and get access anywhere. Now with the reduction in cost of the PC over the last year, two years, three years and the growth of the industry around the world, you are now getting a sort of aftershock of the PC moving into all of the other countries outside of the United States, the Internet PC. And so within two to three years, you will have more people on the Net in Europe than you have in the United States and behind that will be the whole of Asia. You also have a move of the PC into things like cars and much more shrunk mobile form factors. Of course, it's not just the PC. You have other kinds of appliances as well. This appliance here which you may have seen is a RIM Inter@tiveTM pager. It's an interactive pager and it uses the IP infrastructure, the Internet, to send and receive e-mails. This thing has an Intel386 processor in it, so it has the same processing power as the platform that Linus used to develop Linux. And then of course at the same time, you've got the emergence of the Internet-connected cell phone. End of '97 there was something like 200 million cell phones in use. Very, very few of those had Internet access of any kind, probably zero. By 2002, 2003 the projections are there will be 500 million cell phone users and most will come with wireless access protocol and some kind of Internet connectivity. That, of course, tends to increase the problem, the challenges of the second problem, which is sensory overload. And in some ways with all the content that we have and the number of people coming on sending e-mails, it's almost like being under siege. I think Intel is a reasonable example. We've deployed e-mail 100% now for several years, and as the whole organization gets wired up and you get that faster and faster bit rate of communicating and informing each other, the number of mails goes up and up and up, 150, 160 e-mails a day and many people over 200. We need to have better and better client side and server side tools to determine what's important and what isn't importantly. You have the same problem when you go out on the Internet looking for content. And you're all familiar with the huge lists of hits that you get when you do simple searches. That problem is only going to move in one direction as Europe and then Asia also starts to generate vast amounts of content. What I'd like to do now is to take a look at some methods of trying to navigate and make sense of all that information. And I want to start with a local company here which uses Linux called Google. SEAN MALONEY: Okay. So let's take a look at Google. And we should have Randy coming up on stage. Hi, Randy, how is it going? RANDY: Good morning, Sean. SEAN MALONEY: We're going to look at Google and in order to do it, I want to start by doing a couple of searches. We'll start with doing a search on Yahoo. RANDY: Okay. SEAN MALONEY: Not for one second here am I knocking Yahoo. I think they've done an awesome job. My home page is set to Yahoo, and they're doing wonderful things. But what we want to do here is look at what happens if you focus just on the search part of it. So let's go onto Yahoo and look at some cultural stuff. RANDY: Let's start with the Simpsons, television show. SEAN MALONEY: Oops. RANDY: That was not it. SEAN MALONEY: You're leaping ahead here. RANDY: So we're going to do a search on the Simpsons and we get 1,400 products for Simpsons. We get 801 sites for the Simpsons and so on and so on. So information overload. Let's give it a try on Google. RANDY: On Google? Okay. SEAN MALONEY: All of that information we saw there is really, really wonderful information, but let's give it a try. RANDY: Okay. So if we were to click on the Google search, it would act a lot like how Yahoo came up. But if you click on "I'm feeling lucky" it's going to pick the most relevant one and bring it right up. So this is the Simpsons' index page which brings you one click away from -- SEAN MALONEY: Right. You're right where you're right into the Simpsons. RANDY: Mm-hmm. SEAN MALONEY: And there's Shock Wave running on top of Linux. So let's think about a more recent but kind of bizarre cultural reference. RANDY: Well, next to Star Wars, I think that would be something like Austin Powers. SEAN MALONEY: The most hyped movie of the summer. RANDY: Let's not make it easy. Let's do something complicated like just a reference, like "Yeah, Baby" or something like that. SEAN MALONEY: Okay. Let's give it a "Yeah, Baby." RANDY: And click on "I'm feeling lucky" and it brings you right there. SEAN MALONEY: All right. "The Spy Who Shagged Me." Very good. All right. So with that, seeing that we know where we are today, let's go -- RANDY: One more? SEAN MALONEY: -- do a Google search on Linus. RANDY: All right. SEAN MALONEY: This one ought to be really easy; right? RANDY: Sure. Come on. SEAN MALONEY: It's live, folks. Aha, trapped you. You thought -- Okay. Thanks a lot. Of course, those of you know that Linus was actually named after Linus Pauling. Okay. So with that, let's take a look at another example, and this time something that's a little bit more close to the operating system. And I'm going to ask Mandrake , he's already on stage, Mandrake from VA. Hi, Mandrake. MANDRAKE: How you doing? SEAN MALONEY: Great. Now, you've got some rather interesting stuff here with Enlightenment. So maybe you could give us a description of what you've got. MANDRAKE: What we've got running here is a multithreaded system running an X486 3.9 which was just released last week. One of the things that's really neat about this is we can have more than one monitor, we can have as many monitors as we can have video cards in the machine. So we can set up as wide or as tall a desktop as we want to. And that's really great for me, at least, because I don't know about you, but I always have a ton of applications running on my desktop. I have Netscape open and some files open and a shell open and some graphics I've got to look at and some servers I've got to monitor. And then of course you've got to have something pretty to look at, too. SEAN MALONEY: I've got this sinking feeling you may use more computing power than I do. MANDRAKE: Well, I might do that. But like I said, it gives you this really great opportunity to expand your desktop over several different screens and have a single optical display. So as you can see, I've got windows split across all the screens here, so I can take and move one window, move it between one desktop and another desktop, I can move it between monitors seamlessly. So basically, it's fairly new. It's a new technology. But it allows me to have the freedom to have as wide a desktop as I want to. SEAN MALONEY: You have two extremely nice looking flat panels there. MANDRAKE: Not everyone can afford these 21-inch flat screen monitors, so for those who can afford it, you can have several desktops also. So say I have a second workstation or a second works desktop setup I want to have. I can have multiple virtual desktops just like I can have multiple physical desktops also. So I have one desktop set up here and one set up here and I can do all sorts of other things on this one also. SEAN MALONEY: That's great. Thank you very much. Now, how can people find out more about this. MANDRAKE: Well, if you want to find out a little bit more about this stuff, you can go to X486 at www.X486.org or you can go to www.enlightenment.org or come see us at the VA booth. SEAN MALONEY: They said every time we finished a demo we have to throw one of these into the audience. I forgot that one over there. So these are really good Frisbees. Thank you very much. MANDRAKE: Thank you. SEAN MALONEY: I can't resist doing one more demonstration on the user interface, and we just have to sneak this one in somewhere because it's such a fun one. So Randy, maybe we can bring this thing up now. RANDY: Yeah. SEAN MALONEY: If you have kids of a certain age, you're kind of consumed with MP3 fever; right? RANDY: Definitely. SEAN MALONEY: And MP3 chaos. What have you got here? RANDY: It's called the Globe.Com jukebox and it's the front end for a bunch of free software you can find out on the net, as assemble it on your Linux system and run this as your Web front end to it. And just like any jukebox would, you could pick a song and just have it play. SEAN MALONEY: We have the rights to that one because it was in the last Intel advert. And it's a British band, so it's okay. RANDY: What's nice is you can use this to manage your own music so for instance if you have your own music CD, you can walk over to your CD drive, put it in -- SEAN MALONEY: Hold on a second. So obviously what people do now is everyone has got zillions and zillions of CDs, so you build an MP3 jukebox, totally legal, with copies of your CDs. The trouble is maybe you have 200, 300 CDs. You can't sit and type in every title of every track and there no way the PC knows every title of every track. RANDY: Right so what we do is we put the CD in here, the computer will go out over to the Internet to a site called CDDB.com, and using information from the audio CD it will bring up the artist's name and the song title for every song. So simply, you just click on the songs you want to use -- SEAN MALONEY: Right. RANDY: And then go. SEAN MALONEY: I think this is one of the slickest integrated databases and applications I've seen. Saves you an enormous amount of time. Thank you very much. So those are some examples of different things that are being done with user interface and the browser in order to make sense of all the information. The third area of problems I'd like to look at is in the area of instant response. Now, the Internet is, by structure, a very burst-driven environment. And the communications industry has been used to this all through the last 100-odd years, because the telephone system was built around a 24-hour cycle of usage all over the world. The switching structure was put in place to handle these swings. But the Internet is far, far worse than that. It's complicated by two additional factors. One, the marketing campaigns or programs that produce surges, the radio campaigns, the TV ads and so on that produce a big hit. And secondly, on top of that, you then have rich media. For example, the impact of the next World Cup, the soccer World Cup in the evening when half the planet or anyone who has a PC wants to look at streaming video of the goals that got scored, and you have huge spikes of demand. A CEO of one of the largest banks in America was telling me recently about the worries they have on the millennium five or ten minutes after midnight. Going up to midnight nobody will have been accessing the Net. After, everyone will think about the millennium bug. Everyone will try to access their bank accounts to see if their money is still there and you'll get this huge spike. So we need to build an infrastructure that can handle these incredible swings in access. So what I'd like to do now is I'd like to look at some cluster scalability and performance, and I'd like to look at a little demo here and call up John from VA. We've got two machines coming out, you can guess, we'll have a bake off. So what do we have here? JOHN: Two machines, one is a 36 node and one is 19 and we're going to render a PAVR image and we're going to illustrate how we can run processor in parallel. SEAN MALONEY: This is a demonstration of cluster scalability. We have 18 nodes with two processors per node on here, and we've got -- JOHN: Nine modes, SMP nine nodes, two processors each. SEAN MALONEY: It sounds like it will be a fairly simple thing here: JOHN: You're going to win or lose. Do you want to lose? SEAN MALONEY: I'm going to win or lose. Do I want to lose? That's an incredibly existential question. You know, I want to lose. So I'm going to go over here and lose. And you're going to give me the simple idiot's guide instructions JOHN: So all we have to do is one, two, three and go. SEAN MALONEY: Okay. JOHN: One, two -- SEAN MALONEY: A Ha, I jumped the gun. I'm a real loser, I went on two. But I don't think it makes much difference. This is a standard, very high computer sensitive rendering demo. JOHN: Correct. It's a high CPU intensive task and it's why we put it across as many nodes and processors at a single time. SEAN MALONEY: How scalable do you think it is? JOHN: This application itself is very scalable. It depends on the architecture you put it across because you are passing information across a network. In this case its gigabit. But it should scale pretty linearly to about 64 nodes. SEAN MALONEY: Okay. All right. Thank you very much indeed. JOHN: You're welcome. SEAN MALONEY: Thanks. Once again -- hey, John, if people want more information on this. JOHN: Go to VA Linux's booth and we have more information, and also on the Net there's information about parallelism. SEAN MALONEY: Okay. Thanks a lot. Okay. So we're going to come back to performance a little bit later on. We've got one last demonstration left on performance. But I want to now move onto the last of these categories of problems which is total reliability. And again, the challenge that the Internet presents to us is much, much more deep than in the past. In some ways we've been on this kind of constant ratcheting up of requirement going over the past few years. In the 1960s when people ran enterprise applications, 95 percent to 99 percent was kind of okay. That has got ratcheted up and up and up, and now the demand for the server infrastructure on the Internet is kind of really five 9s. Five 9s, as you probably know, is something like 5 minutes downtime per year and there are lots of people who actually are saying we need to go to six 9s to match the telephone reliability. That's down to something like 30 seconds. 30 seconds, I guess you could say might be long enough for a reboot if you had some super quick reboot but it wouldn't be enough to cycle through a memory test on a typical large server with a half a gig or gig of memory. So that really is a challenge. What I'd like to do now is call up San who is going to look at the VA cluster manager and some of the code being embedded there for reliability. How are you doing, San? SAN MEHAT: How are you doing? SEAN MALONEY: What have you got here? SAN MEHAT: What I've got here right now is VA's new VACM cluster management software. We released into the open source community actually today. It's available off of VA's Web site. What it basically does is combine with Intel's intelligent platform management interface which is available on all of therapy server motherboards. SEAN MALONEY: That's IPMI, a really easy to remember acronym; right? SAN MEHAT: It's your company. (Laughter.) SEAN MALONEY: Guilty as charged. SAN MEHAT: Basically what it allows you to do to effect a lot of low level control and monitoring functions that you weren't historically able to do on PCs such as powering a node off, powering it on, hardware reset. All completely remote. Perhaps one of the more interesting features is the ability to actually get a remote console BIOS redirection. So over the network, you can effect changes to your entire BIOS. SEAN MALONEY: So suppose you were the systems administrator on a big ISP with a diverse series of clusters. Give us an example of how you would go about managing that. SAN MEHAT: Okay. Once the VACM software is installed properly, all an administrator need do is scan his network for VACM server. That's synonymous with basically the cluster controller. As you can see, we've already found one which is our little rack over here. Once the scan is complete, I can bring up that node. I'll just drag it from over here. And what you see is a list of all the individual nodes that are in -- that are being controlled by that cluster. Once you've connected, what you can basically do is bring up for any node, bring up a status display, which will tell you a number of things at the low hardware level. The least of which is the power status of the machine, is it on, is it off, the intrusion status, has anyone been inside the case that shouldn't have been inside the case. And the fan status. Are the fans running? Is everything okay? You can also effect in singular node and group operation such as turning the entire node off, the entire cluster off, which I'm do now. SEAN MALONEY: Okay. One second. Let me go over here so I can just check it is going off. Okay. So here is the smaller of the two clusters: Bang, yep, and it's gone all quiet and gone off. And I guess you can switch that back on again. Yep, and it's come back on again. Great. All right. Well, I could probably throw a problem at you by pulling out one of these cables here. SAN MEHAT: Not a problem at all. Once VACM detects the loss of contact with IPMI, it will throw up an alert. There it is. And basically at that time I would give you a call and probably yell at you and ask you to please plug it back in. Once you had plugged it back in, I would be able to determine, after talking with IPMI, that the fault was caused by a loss of AC power. SEAN MALONEY: Okay. So how would you say this is different to doing other types of network management? SAN MEHAT: Normal network management software relies on a type of software to be actually running on the node that's being monitored. With IPMI and VACM, it's completely unobtrusive to the local CPU at all. You don't have to run anything on the target node, nothing at all. The other thing is that the software-based management server, client/server systems that have been historically available, don't give you the ability to effect a lot of low-level hardware type things like the power down, power up and reset without resorting to arcane and expensive solutions like network addressable power supplies and that type of stuff. SEAN MALONEY: So driving it through the hardware means you get access to the BIOS information and so on in the boot sequence. SAN MEHAT: Absolutely and that's perhaps one of the more interesting things to people interested in COLO is the remote accessing features. There's nothing you can do at the physical console of the machine that you cannot do remote with IPMI VACM. SEAN MALONEY: There's obviously a lot of work that's got into this. You've got birds of a feather session later on; right? SAN MEHAT: Yes. We have a room that holds about 150 people and to find out the time and place for that, just come to the VA booth. SEAN MALONEY: Okay. And if people can't get into the room, then they can get move information; right? San they can get more information off the VA Web site or I will be in the booth giving detailed demos, something a bit more intimate. SEAN MALONEY: Brilliant. Great. Now, you told me VA has got some kind of really weird handshake; right? SAN MEHAT: Yeah. SEAN MALONEY: Boom. Okay. SAN MEHAT: Not bad. SEAN MALONEY: Thank you. I'm really worried doing that in case I missed his hand and hit him somewhere else, but then he's Canadian and I'm British so we'd probably get away with it. (Laughter.) SEAN MALONEY: Okay. So I said that we hadn't quite finished with the performance demonstrations, and we've used nearly all of our game show style doors. But we've got one demonstration left. Should I go for the left door or the right door? AUDIENCE: Right. SEAN MALONEY: Right door. Okay. We've got a bunch of right wing people in the room here. Come out, come out, whoever you are. Aha, come on out the front here. ANDREW S. GROVE: Instant availability. SEAN MALONEY: Okay. For those of how don't recognize him, this is Andy Grove who is the chairman of Intel. And you must have just been wandering by. ANDREW S. GROVE: I have been wandering by, listening to all kinds of buzzwords that I couldn't understand. (Laughter.) SEAN MALONEY: Okay. ANDREW S. GROVE: But I'm here to tell you something. SEAN MALONEY: Right. ANDREW S. GROVE: All of this stuff is fueled by a very major change in the business environment, and that change has to do with transactions and business moving into a new medium, a new way of doing things, electronic transactions, electronic connections, much along the lines of everything you've demonstrated here. And I have some data here that shows the portion of the U.S. gross domestic product that purportedly is on the Internet today done by e-commerce. When you look at 1999, you can see that we are approximately one percent, there's a semi log plot -- you know about those semi-logs. SEAN MALONEY: Yeah. I think you told me that one. ANDREW S. GROVE: And one division is a tenfold change. So two divisions correspond to one percent of the U.S. gross domestic product being done on the Internet today. And it's suggested that in three, four years time, we're going to close that gap to within ten percent, to be ten percent of the total gross domestic product being on the Internet. Now, I have a lot of problems with these numbers, and you are the best witness I have to bear me out because one percent where we are today is 100 billion dollars in the U.S., and how much will Intel do in '99 in e-commerce? SEAN MALONEY: We'll do probably 15 to $18 billion. ANDREW S. GROVE: With all due respect -- SEAN MALONEY: We're not 15 percent, that's for sure. ANDREW S. GROVE: We're not 15 percent of the U.S. e-commerce so that number, if anything, is understated. And what that trend suggests, whether it's real or whether it's understated, is an enormous prospective demand for servers. And when you compare the installed base of servers today, which is on the left bar and, and scale it kind of by the seat of our pants, adjusting to the growing demand for the Internet in different parts of the world, different people getting on the Internet, looking at the increasing use of diverse services, some of which you have demonstrated today, but it's only a little portion of it that you've demonstrated today, particularly when you consider the amount of innovation that's taking place on Internet applications, looking at increasingly rich use of data and availability of data on the Internet, and lastly, consider that some slack has to be built into the system because as the dependence on electronic services is going to skyrocket, users' tolerance for down time is going to decrease, and the only way or one of the ways that suppliers of electronic information will have to do to defend themselves from that is to build in some slack. What all this works out to be is the right bar of server demand which suggests that something like in 95 percent of the servers that will be needed by 2005 are still yet to be deployed. And as you know, we are working very hard to develop a next generation product that is meant to fill this role extremely well. That's the IA-64 generation that is likely to do the same improvement in performance and capability as the Intel386 did when you were programming Unix on the Intel286. And we think of it as the future engine for e-commerce. And I think this is a wonderful opportunity for us to give a little bit of a brief update on that. The first instantiation of the IA-64 architecture, the Merced processor, is in the fab, which will yield some silicon in a few weeks' time and we'll either know if it works or it doesn't work. But in the meanwhile, we have developed some very sophisticated simulators for this architecture and we have booted eight operating systems on this processor, one of which is a Linux. And this processor has some features, which are going to be extremely significant for e-commerce application. Of course, the 64-bit addressability will enable people to store large amounts of e-commerce data in memory. The Intel architecture lends itself very well to transaction processing. And the outstanding floating point performance that was designed in the chip lends itself very well to data mining, getting out of information overload you talked about earlier. But I think perhaps it is time to demonstrate what Linux looks like on the IA-64 architecture. SEAN MALONEY: Okay; great. I think the explicitly parallel features are going to be interesting tying in with the cluster demo John gave earlier on. What we have here is the Merced simulator running on this system here and the system is a dual Pentium III processor at 550 MHz. I'm sure you all have one of those in your home; right? With about a half a gigabyte of memory. And we have the Merced simulator, and running on top of the simulator, we have the Linux -- I'm sorry, the Linux IA-64 kernel running on top of it. So this is the first time we've ever shown it. And the work is a result of the Trillian team, which is SGI, HP, Cygnus, Intel, VA have been hosting it, and we're pleased to announce IBM. ANDREW S. GROVE: Right. SEAN MALONEY: I think IBM announced yesterday or this morning that they were part of that effort. So we have here plugged into the Merced simulator, hopefully soon to be running on the silicon, we have IA Linux 64. Now, to test the Linux 64, we have loaded up the -- probably one of the most common applications, which is Apache Web server. So we have basically here a Linux IA-64 Apache Web server, and we'll generate a transaction across to it. This is the first time in public you've seen an IA-64 Web transaction. And there you are. Nothing magic about that, but it served up the transaction and we've gone and done a hit appropriately enough on the Trillian page. So there you are. ANDREW S. GROVE: Well, the only remaining thing that I want to bring in here is that what our plans are with IA-64, starting with Merced and going beyond that, is to bring all the power of the PC economics to it. We'll market it to the world's manufacturers of computers. We will will also evangelize it to the world's software manufacturers, and we hope to bring a very wide variety of choice that lends itself not only to varieties of applications, but to the economic power that this type of a horizontal business arrangement brings to the user, that is what you described earlier, the lowering cost of computing power and continuing with the corresponding increase in power, much in the same way as we have done for the last 25 years. SEAN MALONEY: Okay. Thank you very much. ANDREW S. GROVE: There's only one thing left. There's a lot of people listening to what we're going to do, and some of them may be interested in participating in this -- SEAN MALONEY: I bet a lot of people will be. Okay. So how do you get involved? The Trillian team is going to release the IA-64 kernel source, open source, in Q1 of next year. Realistically that will be a time when we've had time to take a good look at the silicon and get it out there. So that will be released. We'll have a network around the world of application solution centers that we are setting up where you can go in and run your code and test your code if you want. More appropriately for the Linux community, we're also going to make this available, series of Merced servers available hanging on the Net in a number of obvious locations -- S.u.S.E., TurboLinux Caldera, Red Hat, and VA -- we'll put those servers in those places with the welcome support of those organizations and make the systems available across the Net. ANDREW S. GROVE: So people can actually do development, application development and tuning from their homes, from their offices, or from their employer's offices. SEAN MALONEY: Yeah. Obviously the Linux community knows how to exploit the Internet for development incredibly well, so that makes the most sense for the way to go ahead with this. ANDREW S. GROVE: That's good. SEAN MALONEY: Okay. ANDREW S. GROVE: One more item. SEAN MALONEY: One final thing. Those of you who have the entrepreneurial gleam in your eye, about a month ago we announced a venture capital fund called the Intel 64 Fund which Intel announced with a bunch of other companies and the idea of this is we put the money in a big pot and the money will be invested in companies who have interesting, new products that will run on IA-64. And that money, this pretty substantial amount of money, and as we're speaking we've announced this thing and more than 90 percent of the money is available. So if you've got ideas for any really interesting products or you want to do something new and interesting on this architecture using EPIC, the explicitly parallel features inside of it, then please go along to intel64Fund.com or you can send your proposals to proposals@intel.com. ANDREW S. GROVE: Okay, Sean. Can you finish without me? SEAN MALONEY: I think I can probably finish without you. ANDREW S. GROVE: Okay. SEAN MALONEY: Thanks a bunch. (Applause.) SEAN MALONEY: Okay. So, really, in summary, I think that the message here, born to be wired, the Linux community is absolutely born to be wired. We've had great fun over the last couple years working together. The challenges over the next two or three years I think will dwarf what we've seen and I think we'll have an incredible amount of fun. Thank you very much, indeed. (Applause.) * Other names and brands may be claimed as the property of others. |
|