existing constraints into account, don’t talk about incremental change. Start by asking what is the right thing to do.”
One of the things that grew out of that was a deep look at ways to reduce cooling costs. Another aspect was finding the sweet spots for microprocessor design. How much performance do you really need? How do you optimize operations per joule? Those are all long-term explorations that are not going to affect the next generation of x86 hardware.
X: From that focus on data centers, I take it XCG is mainly concerned with the future of cloud services?
DR: Yes and no. There are two kinds of interesting innovations taking place right now. Cloud is one, but there is an enormous amount of ferment in the mobile device world. There could be 50 billion of these interconnected things in a few years, most of which are not computers.
One of the things I’ve learned is that the pendulum swings back and forth. The questions don’t change, but the answers do. Right now the best cloud data center would be this massive thing occupying tens of thousands of square feet. But the pendulum has swung back and forth between centralized and distributed many times before, so one of the things that we are looking at is how do we build “micro data centers,” and how do we reach the other four billion people [in the developing world]. So that we’re not just deploying infrastructure where it’s already available, but so that we can bring information resources to anyone, anywhere.
Maybe I would answer your cloud question this way. I believe there is huge power in local sensors and devices with low-latency response and context awareness. I believe there is also huge value in extracting insights from aggregate data that is consumer- and organization-generated. The fusion of those two things is where the revolution is. The devil is in the details about how we design each of these. Cloud data centers are just one solution point.
X: How is it that one person at Microsoft came to have your two jobs—directing the eXtreme Computing Group and also being a vice president of technology policy and strategy?
DR: It sort of evolved that way. The eXtreme Computing Group is really trying to take a long-term but integrated view of where we think technologies are going, rather than the traditional research view of looking at specific piece of technology. The mission is to step back and say, “If we took this and this and this and put them together, we could make that happen,” rather than just focusing on advancing the state of the art in one area.enough data center stuff already I think.
Every one of those issues raises a counterpart in the policy space. What are the implications of security in the massive cloud infrastructure? That’s already in the news frequently. Craig Mundie was thinking about both of these things, and he needed help, and I was in the right place at the right time.
X: Can you give me a concrete example of how technology research and forecasting and making policy recommendations overlap in your job?
DR: Let’s start with security. Part of what XCG does is look at next-generation hardware standards for trusted platforms—the hardware that supports basic cryptography in microprocessors. If you think about the chain of trust required to run secure software, it begins with the hardware. You want to be able to manage keys and verify that the software that’s running on the hardware is the software you think it is. As we work on next-generation cryptography and mechanisms for managing public and private keys, those technical developments and the software and hardware standards we produce naturally focus attention on the corresponding policy issues.
Say, for example, you’re managing medical records and you want to
