minutes

IBM/UCB Pervasive Computing Meeting 1-9-99

Important Links:

The single overriding message from the meeting was that the collaboration between UCB and IBM around pervasive computing is proving to be extremely effective. It is important. It is in the direction in which new research activities are emerging. It is capturing the imagination of many of the participants and opening up new opens. And, the seed activities started last June have proved very valuable for stimulating this process.

The organization of the meeting was structured to build from direct hands-on interations with students and their initial research projects, followed by more focused faculty and graduate student presentations, followed by discussion on how to continue further advance the collaboration.

The collaboration initiated last June has proved to be very successful - far beyond expectations. At that meeting we asserted that it we wanted to change the culture at Berkeley, not just start a new research project. Although we desired to put small, connected devices in the hands of all researchers in the department, we recognized that the infrastructure would not yet exist to support that successfully. Instead, we elected to seed 150 workpads (along with the 50 or so PalmPilots) into selected subsets of the the research community to start creating that infrastructure. This included the researchers involved in the Ninja and Iceberg projects, which are building a new platform architecture for services that provide intelligence in the infrastructure to support small devices, as well as Millennium, which provides the testbed. James Landay ran an experimental version of the User Interface design class (cs 160) with 60 students all working directly with new interfaces for small devices. This also seeded a number of talented undergraduate. In October, we held a kick off seminar to initiate the discussion of what the Post-PC era might bring and what the core research challenges might be. At that we provided workpads to all of the first year CS graduate students. About 20 additional students joined in with their own ideas. Research was conducted as term projects in at least four graduate courses, including operating systems, computer architecture, and network security. Many of these were excellent - and several of these were featured in the meeting. In addition, the effort has created common ground for several important research projects, including NINJA, IRAM, and MASH. It also altered the strategic plan for the department, with most of the critical areas of hiring taking on a pervasive computing flavor. This activity culminated in a major proposal to DARPA, comprising 14 faculty members and a 21M$ budget, to lay out the direction of experimental systems into the 21st century. (A draft of this was provided at the meeting.)

The presentations and projects are available at the agenda page above.

Several action items emerged, listed here from concrete to broad.

David Culler will work with Paul Gray and Jean-Paul Jacob to map out how to utilize next year's SUR support to further advance this activity. (Immediately following the IBM meeting the ISRG retreat established several of the requirements, which included using at least two devices - one more powerful than the workpad - connectivity options.) Landay's research is posed to harness new crosspads directly. We both need to utilize current, cost-effective devices to build community and assemble prototype of future devices. Both will need connectivity and infrastructure.
A collaboration will be established between the Ninja project and the T-spaces effort at Almaden. We will transfer technology both directions and determine how we might collaborate on further research (including Toby Lehman and Michelle Munson from IBM )
A berkeley team will do the reverse visit to Hawthorn to learn what IBMs directions, accomplishments, and potential devices and interconnect might be. This is hope to yield a broader, higher-level framework for collaboration.
Berkeley will continue building out its pervasive computing infrastructure, with at least one service moving into production use by end of semester and a working connectivty open in key areas of the building. Futhermore, we to build a simulation environment and debugging environment to greatly enhance productivity in building applications.
Make Berkeley a beta-test site of new and emerging IBM technology - HW and SW. ( Paul Gray liked specially the idea of testing replacements or place-holders for Bluetooth-type technology.)
Making Berkeley a "customer" in a FOAK-type study, ie, creating a large experimental Pervasive Computing space in Soda Hall and test technologies, conduct experiments, etc.
Having Berkeley join one of IBM's First of a kind (FOAK) studies with customers (making it a multi-way FOAK) as in, for example, a current study with a supermarket chain in the UK. This will probably hit strong walls of resistance from lawyers and the customer who wants to protect the data and traces of systems usage
Farm out development projects to Berkeley, taking advantage of Berkeley's unique set of skills and environment.
IBM to create a small research facility near Berkeley to jointly do R&D with Berkeley. This idea was labeled "get people closer to Berkeley"

As one of the first steps in implementing these ideas, a group of Berkeley faculty/students will come visit IBM Research
facilities: Almaden and Hawthorne. Jean Paul will host/organize the Almaden visit. Brent Hailpern will organize and host the Hawthorne visit.

David Culler and Eric Brewer are to run a more formal design seminar in post-PC systems, providing a catalyst and source of vision for the post-PC bazaar emerging at Berkeley. Dave Patterson is organizing a faculty friday "great thoughts" meeting.