Roy Want Principal Engineer Intel Research/CTG

Roy Want is a Principal Engineer at Intel, a member of Intel Research/CTG, and leader of the Ubiquity Strategic Research Project (SRP).* He is responsible for exploring long-term strategic research opportunities in the area of Ubiquitous Computing.

Want received his B.A. and Ph.D. in Computer Science from Cambridge University. He is the author or co-author of more than 25 publications in the areas of mobile computing, distributed systems, wireless protocols, multimedia systems and electronic identification. He also holds 39 patents in these areas.

Q1: How did you get involved in ubiquitous computing, and what contributions have you made to this emerging area of research?
A1: In the late 1980s, when I was working at Olivetti Research, I became interested in developing smart telephones. This was the era before cell phones were practical for widespread deployment. I wanted to figure out how to augment the wired PBX telephone systems that were popular at that time.

One feature that users commonly requested was the capability to route incoming telephone calls to a location closest to the person for whom they were intended. To solve this problem, I developed the Active Badge a wearable device that wirelessly communicates the wearer's identity to nodes in a network of sensors deployed throughout a building. Our team built services that provided location information, based on the sensor data, to applications that knew how to control call delivery within the telephone system. As we later discovered, these services also opened up the opportunity to build context-aware mobile computing applications.

In 1991, I joined Xerox PARC's Ubiquitous Computing program. Among other research efforts, I led the PARCTAB project and built one of the first context-aware computers. Our goal was to explore the capabilities and impact of mobile computers in an office setting. The system consisted of palm-sized computers that communicated wirelessly with workstation-based applications. The handheld device was unique in that it was aware of the room it operated in and other people and devices in the locality. As a result, the applications it supported could organize the data they displayed to take into account the context of use. For example, if you wanted to print a document, the nearest printer available at that location would be shown as the recommended printer.

At Intel, I've continued to pursue my interest in ubiquitous computing. My research currently focuses on the personal server concept, which I'm developing along with my team as a means of enhancing the mobile computing experience.

Q2: What is a personal server, and how did you develop the concept for this device?
A2: One of the main reasons handheld mobile computers are severely handicapped is the limitation imposed by the small display and keyboard inherent in their design. In looking at ways to improve mobile computing, we focused on three technologies, which in combination, offer a potential solution: high-density, small volume storage; low-power, high-performance processors, such as StrongARM and XScale; and standardized, high-bandwidth radios, such as Bluetooth.

The personal server represents the integration of these three technologies. It's a small, mobile device that eventually will hold most of the data you use from day to day. It has none of the standard physical input/output capabilities no keyboard, buttons or display. For this purpose, it relies on the computing infrastructure that happens to be in the locality and short-range wireless-connections to make use of these resources.

A simple way to think about the concept is to ask this question: What makes your PC your PC? The answer is that it's just the hard disk, which contains your data; the rest is simply the access device. Using the personal server concept, we're virtualizing the hard disk through a wireless connection to whatever computing device is nearby and available.

Back To Top

Q3: How would I use the device? Can you provide an example?
A3: Suppose you're traveling on business and carrying your personal server. You could sit down in front of any PC that has the necessary wireless capability and a small icon representing you would appear in a corner of the screen. When you clicked on the icon, it would expand to a much larger window, perhaps with a set of documents and applications that you are currently using, and you could access them in the same way as if you were sitting at your own computer.

One model we're exploring is making the personal server very much like a Web server. The personal server system would automatically launch a browser on the PC you were accessing, and within the browser pages you would see iconic representation of your applications and data.

Q4: What are the advantages of the personal server over other mobile computing models?
A4: One advantage is convenience. In our design concept, the personal server is small about the size of a deck of playing cards so it would be easy to carry with you wherever you go. It could interact with any nearby interface, such as a desktop PC or an information kiosk. It would not have to be directly accessible to the user, so it could be in the bottom of a purse, in your pocket, or perhaps in the future be part of your shoe. As long as you're in the vicinity of some computing infrastructure, you would be able to access your data.

The personal server also overcomes the display limitations of other mobile devices. Although it is possible to access personal data through existing small-screen mobile devices such as PDAs and cell phones, it is generally inconvenient, awkward, and slow. The personal server has no display, but instead uses large-screen displays in the local environment.

A wireless PDA or cell phone could also be used as the target display for the personal server. It would have the small screen limitation, but if nothing else is available it would have to do. In this way we also introduce the concept of scrap computers: when PDAs become inexpensive enough that they are left scattered around, in the same way we treat pens and paper today, you could pick up any of these devices and through the personal server it would, by association, become your device accessing your data.

Another advantage is control over your data. With the personal server, you wouldn't have to rely on Internet access to retrieve your data. And because your data travels with you, you know that it is secure and you readily have access to it.

Suppose you are traveling to a conference to give a presentation in front of, say, 300 people. Would you want to rely on Internet access to retrieve your slides from a remote server at your work site? The connection might not be available, the server could be down, and there may be firewall or other permission issues. Even if the connection can be made, it might have latency issues and perform very sluggishly for the audience. Most people would choose to bring their own laptop with a copy of the presentation, to guarantee availability. Some might even carry an additional set of acetate slides as a backup. The personal server concept would provide a level of confidence close to the latter example.

Q5: Outside the office, what sorts of applications do you envision for the device?
A5: There are numerous potential applications. One possibility is that the world could be augmented with information beacons that transmit information to your personal server. As you walk by Starbucks, you might acquire a coupon for a discount latte. As you pass restaurants or storefronts, menus and sales information could be captured. In a train station, you might automatically receive a copy of the timetable.

The preferences for the type of information you are interested in receiving, and the information you would like to export, could also be stored on your personal server to control its operation. An export service is one that is applied to other nearby devices. For instance, you might wish to have your familiar speed dial numbers applied to any cell phone you are currently using. In an automobile, you could automatically apply your preference for music or temperature, or make seat adjustments. Essentially, any rental car could have its user settings be the same as your own car's settings.

Another application we've considered is personal health monitoring. In the case of someone who has a diabetic condition, the personal server could connect to cardiac and blood glucose monitors, acquiring data from sensors distributed on the person's body. Then, when the person moves close to a PC connected to the Internet, the data could be automatically downloaded for processing.

Once you find utility in carrying a personal server for its mainstream uses, the secondary applications I've just described may be value-added products for a new business.

Q7: What aspects of the personal server are you working on within Intel, and what is the status of your research?
A7: We're working on both the hardware and the software systems required to support the concept. Currently we're putting together some of the nascent protocol stacks and software components that are being established by our industry. By combining these components in novel configurations, we can find their strengths and limitations, and recommend how the standards can be improved.

Bluetooth is one of the components we're looking at, and UPnP is another. Since UPnP is included in Windows XP, this is an important part of our PC implementation. By using Bluetooth and the PAN profile, we are able to establish a TCP/IP connection between a PC and a personal server and then discover UPnP services on the personal server.

We have built prototypes of the device and have demonstrated them at a number of events, including the Intel Developer Forum this spring. You can place the prototype device near a PC that is wirelessly enabled and automatically launch a browser that connects to a mini-website on your personal server. You can then view data in the same way you would view data on any conventional Internet Web site.

Back To Top

Q8: When can we expect to see personal servers in the marketplace? What needs to happen before they are viable?
A8: There's a possibility that we could see early versions of the device on the market within three or four years. We still have some challenges to address, such as the availability of high-density storage. Higher storage density provides more utility for the personal server concept. With storage density roughly doubling each year, we should have appropriate storage devices available for our purposes within three years or so.

In order to create a product and market around the personal server concept, we also need a wider deployment of wireless infrastructure. We're seeing Bluetooth begin to roll out, and we're seeing 802.11 and its various incarnations becoming established. Within three years, the clear winners will begin to emerge and there will be wider deployment of whatever standards are adopted.

The other defining variable in the success equation is power. To make the personal server attractive, it must be "always on" but assume a low-power state when not in use. The device must have a suitable power density in its power source, so that it doesn't have to be charged on an hourly basis, but instead could be conveniently charged on a more practical interval of, say, one week. Within three years, some of the power issues will have improved, through advances in processor and wireless technology and potentially in micro-fuel cell technology, which could lead to significant increases in battery energy densities.

Q9: What are the key roles that Intel will play in transforming the personal server from concept to reality?
A9: Intel will provide the high-performance, low-power processors that are required to build personal servers. We are also at the forefront of developing and refining the standards to support the concept, such as UPnP, Bluetooth, and 802.11x.

Finally, Intel is pioneering some of the new storage technologies to increase storage density. If these technologies are successful, it will result in an order of magnitude leap in memory capacity. Not only will this make the personal server a viable concept but it will inspire a whole range of additional new products.

The Ubiquity Strategic Research Project (SRP) is a collaboration with Gunner Danneels, Muthu Kumar and Peter Adamson of the Emerging Platforms Lab (EPL); Jim Kardach and Graham Kirby of the Mobile Platforms Group (MPG); and Gaetano Borriello, director of Intel Research Seattle. The project's core team members are Roy Want, Trevor Pering, Murali Sundar, John Light and Alex Nguyen.

In addition to basic research, the team has a charter to put Intel on the map in this area of research through publications in premier conferences, workshops, technical program committees and through top-tier university collaborations. Projects such as Ubiquity are well suited to making connections between Intel business units and the new Intel University Labs, recently established by Intel Research, such as those in Seattle, Berkeley and Pittsburgh (associated with the University of Washington, UC-Berkeley, and Carnegie-Mellon University, respectively). SRPs can provide the missing link to connect a University Lab with business groups, other organizations such as the Emerging Platforms Lab (EPL); and engineering resources, such as core Intel technologies. For this reason, the Seattle and Berkeley Labs, and Ubiquity already have a strong connection.

The Ubiquity project was launched nearly a year ago. To date, the project team has filed four patents and published two journal articles, with three more in the pipeline.

Back To Top

Back Home