Although it's among the market leaders in the low-end server business, Sun Microsystems has in recent years been known much more for its success at the high end. In this interview with Computerworld, Neil Knox, executive vice president of Sun's volume systems product group, talked about his company's recent blade-server product launch, its emerging throughput computing strategy and Linux.
Q: Why has Sun moved to a quarterly release schedule for its products? Why should users care?
It makes their lives a lot easier. It forces us to pull all the products together in a coordinated manner. So the customers ... are able to plan the implementation of new [and] refreshed products, as well as updates to their data centers. By doing a quarterly launch, we are really showing them a total systems play rather than stand-alone product announcements, which they have to pull together to make sense out of.
Q: So how has this affected Sun's low-end technology and product direction?
I think it is no longer just a question of providing low-cost hardware. That is an old, single-element strategy. With the blade farm that we are able to put into the shelf and the rack, the customer is now looking for productivity tools and software systems that enable them to harness and package the power of this computer.
So that is where the design of the Sun Fire blade platform really comes into play. We gave it a mix-and-match-and-manage environment so that [customers] can pick their architectures. If they want SPARC/Solaris or Solaris/x86 or Linux/x86 environments, they have that choice.
It is very much building on the success that we've had over the last few years. We are raising the bar and saying it is not just about hardware performance, especially at the low end. It would typically take a user about two days to take 40 servers and plug [them] into the data center, rack them, stack them, bring them up, plug them in and burn them in and make them productive on the network. Now you can do that, we believe, in the design of the Sun Fire blade platform in about two hours.
Q: Some Wintel vendors argue that it's all about cheap processing power at the low end. The game is picking up.
It is not just about cheap compute power. It is about the ability to manage that capacity and manage that power and apply it to the user and to improve their ability to deploy new systems, applications and resources. That is why provisioning and the software we have got in the N1 architecture that we are rolling out is going to be such a major play in this environment.
Q: What can users expect to see from Sun in this regard?
I think you are going to see us continue to drive the capability to take these building blocks and optimize them for specific workloads. You are going to see us make it easier and easier for customers to manage pretty massive amounts of compute power seamlessly and efficiently. You are going to see us add more and more power in the blade format in this space. You will also see us drive throughput computing ... allowing us to put multiple streams of compute power into [a] single chip and really increasing by an order of magnitude the price/performance at the low end.
Q: What really is throughput computing, and why does it matter for users?
It is really about price and application performance. Instead of a one-lane highway going through a CPU, if you are able to have a massive speedway all with parallel data going through the CPU at the same time, you can get a picture of the compute power that we are [going] to put on a chip. So if you are a heavy transaction-processing TPC-C type environment, and you have got a CPU that can run multiple streams of data at the very same time in parallel, you have got yourself a next-generation break-out-of-the-box computer that is going to be totally awesome in price/performance.
Q: But will commercial applications need to be somehow tweaked to take advantage of this sort of parallelization?
We don't believe so. The users and the [independent software vendors] see the paradigm changing, and they are going to drive the model in the environment to take benefit of the price/performance. You will see [products] next year. They will be rolled out in a very aggressive format in the next two years.
Q: Your detractors argue that Sun is doing this to divert attention from the fact that SPARC today doesn't quite match up to the competition.
They have all given up. We are the industry-leading 64-bit architecture today. We are looking for breakout strategies as to where we are going to move the model next, whether it is Java or NFS [Network File System]. I mean, that is what we do here. We just don't repackage somebody else's technology. We do actually sit in little rooms and think up stuff as to what our customers are looking for and how they are going to drive the computing model. It is our own intellectual property.
Q: What is Sun's Linux strategy?
Every time someone speaks to Sun, there seems to be some subtle difference. We do Linux as Unix. It is not a competitor to us, it is a friend. Anybody that buys or installs Linux means they are not installing or buying or dealing with our friend in the Northwest. Anything that says, "No, I am not going to spend my money with the monopoly known as the 'M-word,'" is good news to us.
I think you will see us over a period of time go more and more to a general-purpose Linux environment. That is why we announced the blades with the mix, match and manage environment with Solaris and Linux. Some people think we are not serious. But we are deadly serious, because if you look at N1 and the software that's being developed there, it is all going to run on Solaris and Linux. ... The reason we announced Linux systems -- x86 systems, I may add -- was again to say to the customer base that we are very serious about Linux, and we believe that it is a fair playing field at Sun. We even announced Mad Hatter, which is Linux for the enterprise desktop.
Q: Is Sun struggling somewhat on whether to push Solaris or Linux at the low end?
It is not a struggle. It is a natural evolution of the customer environment. We didn't sell x86 systems when we developed Solaris for x86. When we tried to pull away from Solaris x86, we got tremendous push back from customers. And so we listened to them, and now we have rolled out low-cost 32-bit systems. So if a customer wants to take advantage of a 32-bit environment and roll it out on the network and edge, they have Solaris. Customers are coming to us and saying, "You've got to continue providing us with the environment that we have become used to on Solaris." On the other hand, they may also be deploying Linux somewhere else, and they want to be able to get that from Sun. [Customers] are looking for a selection of architectures to fit their needs. That is why we offer these architectures. I don't believe we are struggling at all. I think we have a very simple and very effective architecture strategy.
Q: So what is driving Sun's recent low-end thrust after years of pushing high-end stuff?
I'm on the second floor. Every time I hear this question, I want to jump out of the window. Where do you think we came from? We were a low-end shop when we started. Yes, we were supersuccessful with the Cray acquisition, and we have introduced killer products like the E10K, 12K and 15K [servers]. In the meantime, and I'm not looking for sympathy here, the little volume server business has continued to be a major element in the company. If you look at what we did with the 880, we got an eight-way powerhouse server there for under $50K. We are still the industry's leading 1U [1.75 in.] server. My No. 1 volume product is still under US$2,000. Year over year, I am the fastest-growing company in the $100K space. I outperformed every competitor in Q4 of last year. Despite all the press and all the ads that IBM runs, I am still No. 1 in that space. I think our success at the high end has kind of overshadowed [us], and I think we have not done too stellar a job of driving our success in the marketplace down to the media.