EECS Systems Research in the Post-PC Era
David Culler
U.C. Berkeley
EECS (ILP) Conference
Feb 18, 1999
http://postPC.cs.berkeley.edu
2/18/99 EECS PostPC 2
Format of the Session
• Morning Highlight Talk - Dave Patterson– Computer Architecture and the Infrastructure
• Systems Research Agenda - David Culler
• Ninja Service Platform Architecture - Steven Gribble
– Push scalable services into the infrastructure
• Security in a Pervasive Computing Environment - Mike Chen
– distributed due to limits on power and trust
• Comfortable pace & Lots of Discussion
2/18/99 EECS PostPC 3
The Emerging Platform Pyramid
SuperComputers
SuperServers
Departmental Servers
Workstations
Personal Computers100 Millions
Millions
Thousands
Workstations
Small DevicesBillions
2/18/99 EECS PostPC 5
Natural Tides of Innovation
Time
Integration
Innovation
Log R
Mainframe
Minicomputer
Personal ComputerWorkstationServer
2/99
2/18/99 EECS PostPC 6
Historical Perspective
• New eras of computing start when the previous era is so strong it is hard to imagine that things could ever be different
– mainframe -> mini
– mini -> workstation -> PC
– PC -> ???
• It is always smaller than what came before.
• Most think of the new technology as “just a toy”
• The new dominant use was almost completely absent before.
• So where are we headed in the post-PC era?
2/18/99 EECS PostPC 7
Away from the “average device”
• Powerful, personal capabilities from specialized devices– small, highly mobile or embedded in the environment
• Intelligence + immense storage and processing in the infrastructure
• Everything connected
Laptops, Desktops
Devices
2/18/99 EECS PostPC 8
Complement to industry efforts
• Get maximum number of applications first– 1990 PC capality in handheld device
– microkernel port of Unix or Windows
– emulate vast API
• Mobile extension of dedicated PC– take short excursion and synch
• Success of the Palm Pilot with primitive OS and split application model is significant
– it’s the approach, not the technical superiority
2/18/99 EECS PostPC 9
Example - PDA scope
• http://www.21store.co.uk/pdantic/pdachart.htm
2/18/99 EECS PostPC 10
Rich set of new challenges
• Natural, high-content user interfaces
• Sensors, actuators, display, speech
• devices, small OS, low power
• massive distributed system
• “Middleware”
• Security, privacy, content
• Networking
• Software engineering
• Administration and management
• Extraction of knowledge from activities
2/18/99 EECS PostPC 11
Future Internet-Scale Systems
• ~10 Billion of Information Appliances
• ~100 Million of Stationary Computers
• ~Million Scalable Servers
2/18/99 EECS PostPC 12
Natural Convergence at the Extremes
• “Internet-Scale” => system reaches “everywhere”– small devices will be what is “wherever”
– powerful servers is where is all goes
• Scalability, efficiency, simplicity, availability, adaptation
– commonality in design goals and technology
– federated systems
• The breakthrough ahead is pervasive devices + communication
2/18/99 EECS PostPC 13
Seeds sewn in many projects
• Devices - Infopad, IRAM
• Scalable Servers - NOW, Millennium
• Storage - Tertiary Disk, Istore, Aetherstore
• Sensors and Actuators - BSAC
• Connectivity - BWRC
• Transcoding Services - Wingman, Mediaboard
• Platform Architecture - Ninja
• Computing/Telephony Integration - Iceberg
• Programming Enviornments and Tools
• User interfaces - Notepals
2/18/99 EECS PostPC 14
A Radical Experiment
• What we need is not just a new research project, but a new “computing culture”
=> Build a department-wide, universal wireless PDA infrastructure and a community to take it forward
• Initial Seed Fall 98 with IBM– 150+ IBM workpads + lots of cradles + IR + ???
• Initial community– Ninja, ICEBERG, MASH grad students
– Senior UI Class (CS 160)
– All interested 1st year CS grads (CS 252, 261, 262 projects)
– Fill out based on interest, talent and availability
=> “ask a good question and get yours” seminar
2/18/99 EECS PostPC 15
Fall’98 Project Excerpts
• E-Commerce and Security– Pay-Per-Use Services on the Palm Computing Platform (Mike Chen, Andrew Geweke)
– Secure Email Infrastructure for PDAs (Hoon Kang, Rob von Behren)
– SyncAnywhere - Secure Network HotSync (Mike Chen, Helen Wang)
• Groupware– Kiretsu - Ninja Instant Messaging Service (Matt Welsh, Steve Gribble)
– The MASH MediaPad - Shared Electronic Whiteboard for the PalmPilot (Yatin Chawathe)
– NotePals - Lightweight Meeting Support Using PDAs (Richard Davis)
– OSKI - Open Shared Kalendaring Infrastructure (Jason Hong, Brad Morrey, Mark Newman)
• OS and Communications– PalmRouter - Networking Sporadically Connected Devices (Andras Ferencz, Robert
Szewczyk)
• Numerous Architecture Studies
• Excellent UI Projects– Ink Chat, Nutrition/Excercise Tracker, Rendezvous - Meeting Scheduler
2/18/99 EECS PostPC 16
Some Lessons
• Communication is enabling– low-power wireless needs to be like IP
• Virtual Environment is important– Devices connect “into the infrastructure”
» Network HotSync, groupware, centralized e-mail
=> Need lean, clean communication substrate
• “User Service” is fundamental– not just profile and customization info
– routing point for security
• Much room for improvement in devices– trade BW for compute or storage
• Development effort is the limiting factor– OSKI: 1 person for infrastructure, 2 for WorkPad
=> need complete distributed system debugging and simulation environment
2/18/99 EECS PostPC 17
Momentum Building
• Millennium provides large-scale testbed
• Ninja architecture allows developers to “Push Services into the Infrastructure”
– scalable, available, customizable
– real services deployed and used in Spring 1999
Gigabit Ethernet
PDAs Cell PhonesFuture Devices
WirelessInfrastructure
DesktopPCs
Servers
Clusters
Massive Cluster
2/18/99 EECS PostPC 18
Emerging Agenda
• Endeavor Expedition (14 Faculty)– extend pervasive computing view to “oceanic” proportions
– massive, fluid data storage
– devices everywhere
– fluid software
– streaming data management
– automatic management
– social networking
• Pervasive computing “stamp” on strategic plan
• Causing us to rethink what we need in our environment
2/18/99 EECS PostPC 19
University/Industry Roles & Collaboration
• Bold, Rich PostPC Agenda Emerging– Pervasive ‘stamp’ on strategic plan
• New balance of expertise and technology between industry and university– devices, components, networks, applications, users
– foundations for the future vs TTM
• New roles and relationships in collaboration– how do we share space, environment, culture, not just technology
• Fundamentally new demands on the research space– ability to deploy smart spaces on a large scale
– new modes of human interaction
• It’s not just what we build, but how we use it