Feature Story
|
Past, present, future.Greg Papadopoulos, Sun's executive vice president and CTO, sketches the landscape of network computing. 24.Dec.02--Probably the most interesting thing that's happened in technology over the last 10 years--and it looks like it will continue over the next decade--is the expansion of the network. The network is growing faster than computing itself. Cell phones, PDAs, your car--all these items increasingly have network connectivity. Retail stores are interested in attaching RFID--or AutoID--tags to the items they sell, which provide a network presence to something as simple as a package of razor blades. Companies like Gillette have a keen interest in this, because it can remake their distribution chains while addressing issues such as razor blades "wandering away" from stores. The same set of base technology improvements that speeds up computers and reduces cost over time also reduces the cost of RFID tags to the point where they end up being ubiquitous. The implication is that the growth of the network is still on an exponential curve, which you can expect to increase--going from hundreds of millions of things on the network today to billions and even trillions of things. The increasing reach of the network and the exploding number of things on the network--and the resulting pressure on the infrastructure--is central to Sun Microsystems, Inc.'s systems direction and to the future of network computing.
The key concerns are cost and complexity. Any company that is switching from delivering software as a shrink-wrapped product to creating services that run over the network and touching its customers wherever they are knows a big leap is necessary. It's a big leap for developers, who are saying, "I used to write software for a single computer; now I have to write software for this network of computers, and I need to come up with an expression that allows me to serve millions of users potentially at the same time." That's what Sun ONE, the Sun Open Net Environment platform, is all about. At the same time, the people responsible for the administration or operations of the infrastructure not only have an environment that is increasingly operational around the clock and has to be responsive in real time, but also must address the scale of the Internet. That problem was left largely unaddressed in computing, but it is among the top concerns of customers. In a typical IT budget, 10 percent goes for acquiring the base hardware and software; 10 percent goes for the applications that run on those systems; and 80 percent goes for the operations, consulting, and services that attempt to make it all work. At Sun, our strategy is not to look at the 80 percent and say, "How can we, a services company, get a piece of the 80 percent?" We want to attack the problem in a different way, starting with the question: "Why is it all so complex, and how can we, as a dyed-in-the-wool systems company, apply our engineering skills to make systems simpler and more efficient and deliver network-scale services?" There's a fork in the road for companies now. Clearly IBM, with its acquisition of PricewaterhouseCoopers (PWC), is saying, "We're going to handle the complexity that has emerged by employing an army of consultants who will take the problem off the customer's plate." But that path doesn't address the inherent concerns about efficiency or complexity. The fork we're taking is to aggressively go after the engineering of the complexity itself and construct systems that operate at network scale. Our approach, which we have mastered over the past half dozen years in our very large systems, is evident in the Sun Enterprise 10000 servers and the Sun Fire 15K servers. Both feature dynamic domains, or partitioning, which divide applications or services over more than 100 processors in our large systems. The machine then moves resources around as a particular application demands more resources, while another needs fewer. Without rebooting or changing anything, we can move processors from one domain of the machine into another, purely electronically and dynamically. So a Sun Fire server system administrator, for example, sees a hundred processors and hundreds of gigabytes of memory as one administration point. But the complexity of administering a 100-processor system is not 100 times the complexity of administering a single-processor system. It may be four times as complex. That's what we want to see at the next level of the network's scale. As you add more servers and storage elements, the complexity of managing them does not grow linearly. That's the essence of an architecture we've code-named N1. We're approaching complexity from our strengths, which is why we're treating it as an operating system problem. The idea is to apply our discipline and knowledge about how to make the Solaris Operating Environment understand environments as complex as a Sun Fire 15K server yet hide the complexity from an administrator. The next challenge is to take our knowledge of the software stack--essentially the network computing infrastructure--and build in a network-scale operating environment in which people can run hundreds of services that can be dynamically provisioned. These accomplishments and objectives define a transition for what we build. We believe that we are leading the way to the reinvention of computer systems, something that historically happens about once a decade. The 1960s brought the mainframe; in the '70s, the minicomputer arrived; the PC and workstation defined the '80s; and the the SMP (symmetric multiprocessing) server revolution was the transforming development of the '90s. Sun profited from its product focus and open architectural leadership with workstations and servers. In the 2000s, we are constructing "network-scaled computers" to provide the foundational processing and storage infrastructure for the evolution of the network. In that effort, N1 is the architecture and operating system; the Sun ONE platform is the developer environment. N1 represents Sun's long-term vision. It's how we intend to add value for our customers who need someone to help them manage the complexity of delivering services on demand. With N1, we're doing that--and laying the foundation for the next 20 years of Sun Microsystems. | ||||||||||||
 | | | |||||||||||
