Sun Microsystems recently unveiled its Network Computing strategy in the pursuit of its N1 vision. Mark Jones, Technical Director Tom Yager, and Editor at Large Paul Krill sat down with CEO Scott McNealy to discuss blade servers and the impact of its datacenter strategy.
McNealy: The biggest strategic whiff of my career was when we organized the company back in the early '90s around the horizontal model and I had Sun Microelectronics to compete against Intel, SunSoft to compete against IBM, JavaSoft to compete against Microsoft, SunExpress to compete against Dell, SunService to compete against IBM Global Services. We're still undoing that. It really got us away from Sun Microsystems. About five or six years [ago], Ed [Zander] and I looked at each other and said, "This just ain't right." So [we] started dismantling this thing piece by piece, and it culminated into Ed [managing] himself out of a job. I needed a COO to be managing all these businesses but there's only one P&L [profit and loss] here. I run the NC [network computing] P&L. It's a very simple company now, and [our recent] announcement was a very simple announcement: "This is NC '03 Q1." It was Vignette that made me understand how powerful this quarterly announcement model is. They write to all those different piecesand they said they never knew what configuration at Sun to certify to and to target their release to. So they had to pick something really old because they had to get the lowest common denominator. Now they can target this new environment and know that it's all been tested and all there.
McNealy: I think this is very powerful. It's going to make it easier internally to manage because we're going to get on a 13-week cadence. Week 1 we'll do pricing, Week 2 we'll do the collateral, Week 3 we'll do training, Week 4 we'll do press and analysts briefings, Week 5 we'll do run-throughs, Week 6 we'll do an announcement, etc. We're going to coordinate the announcement weeks around things like JavaOne, Sun Networks, the analyst conference -- different big events -- so we'll have four big announcement weeks per year. The customer will know what to buy. The hardware and software vendors will know what to write to. The hardware vendors may decide to do Q3 every year and skip the other three because hardware has a longer shelf life. It's going to allow everybody to build a cadence around what we do.
McNealy: Hey, if you miss the train, you got to wait another quarter. They're already scrambling to get on the Q2 train because there was some stuff that didn't get announced this quarter.
McNealy: I guarantee it'll be there in Q2.
McNealy: You know, Sparc Solaris is not a new server, Solaris on x86 is not a new server, Linux on x86 is not a new server, storage is not a new server, load balancing is not a new server -- [they're] just a new form factor. The blade form factor has some advantages. It allows third parties to integrate. So we're going to get a huge [advantage] and we're ahead. We've published the specs, we've got the right format.
McNealy: You're on the right issue and, in fact, nobody's asking this question. There are some constraints with it. There are some advantages. You get third-party help, you get a consistent architecture, you get very granular additions. But that spec has to stay firm for a long time, so you can't update the backplane as often. You can't update the form factor as often, so that tends to coagulate.
McNealy: We have to do both. The Intelligent Shelf is a bigger announcement rather than the blades, and so the Intelligent Shelf might get upgraded in NC '04 Q3, or whatever. I'm just pulling the date out of the air. But we're only going to do the shelf every year-and-a-half or three years or whatever. At some point you just say, "I've got to do a new form factor."
McNealy: I think Cisco will [be] building blades for our environment. And Cisco tends to do a lot of networking outside of the switch room. We're going to probably do a little more of the networking inside the switch room. They might try to come into the switch room -- good luck. You've got to do storage and networking and servers and the software -- all the rest of it. You're not going to see us try and do switching in the control room, in the [wiring] closets. We're not out in the wiring closets, we're in the datacenter.
McNealy: We're using lots of Java code in N1 and people will be writing Java code to profile their components. If you've got a Cisco router [or] a NAS box, we'll have a Java development tool that will allow you to write the profile and write the agent stuff so that it connects and works. Fundamentally, Java's a way of getting the job done.
McNealy: It's not even a component, it's just a style of creating components in the network. SunONE and Java and XML are the programming environment for people who develop servers, whether it be a mail or a transaction or mining or electronic-commerce or a monitoring or video-streaming [server]. We tell you to write to the Java virtue machine -- you write to J2E, J2SE, J2ME, or Java card because then you get the write [once] to run anywhere and it doesn't matter whether it's run on Wintel, Intel, Sparc Solaris, mainframe, or whatever. You also get the common security model from Smartcard all the way until you get a common provisioning, a common development architecture, a common management architecture. So clearly, write to SunONE. If you write to .Net, you don't run on the mainframe. If you write to .Net you don't run on the Smartcard. If you write to .Net, you don't run on my Java phone. If you write to .Net, you don't run on my Macintosh. If you write to .Net, you don't run on my Sun Ray (thin client). If you write to Java, you run on all of those and a Windows environment. That's why we're telling people to write to Java.
McNealy: Absolutely. We don't want to be doing application integration. [We] don't want to be doing Siebel implementations and that sort of thing. We don't want to be necessarily managing PCs, hands-on. We don't necessarily want to be loading stuff into the directory -- we want your employees to be loading things in the directory and running stuff on the apps server, but we'll upgrade the apps server to the next version of the NC. We'll do more managed services from that perspective, but with our employees on our equipment mainly. We will, if the customer requests it, be the general contractor and we'll hire Accenture to do the business process re-engineering. We'll hire Cable & Wireless to be the hoster. We'll hire somebody to manage the PCs for you. We'll do that outsourcing thing but we'll go out and pick best-of-breed, as opposed to in-bred. IBM Global Services is almost all in-bred.