Attending a briefing with Greg Papadopoulos, Sun Microsystems Inc.’s CTO, is like stepping into an MIT classroom. He loves to talk about where we’ve been, where we’re going and why.
A question he gets a lot today is: Is innovation dead? In a word, no. “It is a far more interesting time in R&D than two years ago,” he says. “In hypergrowth you’re focused strictly on tactics. How do I make more stuff? Today, you get to think about the important problems and where we are going to be in the next two, four, five years.”
We’re at the beginning of a phase-change in computing, he says. If the 1980s were all about the PC and the 1990s were about the rise of symmetric multiprocessing servers, this decade we’ll witness the arrival of what he calls the network-scaled computer.
Pushing us toward this future is an 80/20 rule. Eighty percent of IS dollars goes to operations today and only 20 percent goes to capital expenditures, Papadopoulos says. Fifteen years ago it was more like 50/50. We’ve created complex infrastructures that cost too much to maintain.
Another problem is the infrastructure is underutilized. Systems are often running at 15 percent to 20 percent of capacity, he says, while the ideal would be up around 80 percent.
What we need to do, Papadopoulos says, is get beyond the habit of dedicating systems to applications. We need to be able to virtualize computing resources and use them as needed, just like we’re starting to do with storage.
He drew a diagram of a tiered system architecture with firewalls/load balancers at the front feeding Web servers that are tied to application servers, which are ultimately linked to servers and legacy systems that share a back-end storage-area network.
To treat this collage as a network-scaled computer that will accommodate the needs of the moment, you’ll need code on all of the piece parts and a control plane to orchestrate the efforts.
On the software side, Papadopoulos says the shift to Web services should help reduce operations costs by making developers more productive. Instead of building to operating systems, developers increasingly are building to Web service components that sit on top of the operating system, such as application servers, directories, databases and messaging tools.
While it will take years for the development community to completely shift to this model, Papadopoulos says all computing is going this way.
A compelling vision. But the question is, how much will it cost you to get there?