Welcome to the WeBS Retreat David Culler Computer Science Division U.C. Berkeley Intel Research @ Berkeley www.cs.berkeley.edu/~culler.
Download ReportTranscript Welcome to the WeBS Retreat David Culler Computer Science Division U.C. Berkeley Intel Research @ Berkeley www.cs.berkeley.edu/~culler.
Welcome to the WeBS Retreat David Culler Computer Science Division U.C. Berkeley Intel Research @ Berkeley www.cs.berkeley.edu/~culler Outline • What retreats are about • Introductions • Webs Technology Push • Webs application opportunities • Where we are now • Research Challenges • Overview of the Agenda 1/12/2002 WEBS Retreat 2 What retreats are about • 6 month project checkpoint – milestones, accomplishments, shortfalls – course correction • Students refine communication and investigation skills – interested benign audience, lots of feedback • In depth exchange with industrial collaborators – discussion and feedback – close with feedback session • Build team and cement connections 1/12/2002 WEBS Retreat 3 Introductions • the start of a 3-day discussion... 1/12/2002 WEBS Retreat 4 Technology Push • CMOS advance not just Moore’s law – miniaturization, lower cost, lower power, complete systems • MEMS bringing rich array of cheap, tiny sensors – tiny actuators too • Communication – low-power radios, optical, ... • Power – Solar, vibrational, parasitic, ... mixer PLL LNA • => microscopic Processing, Storage, Communication and interaction with the physical world • Can foresee computational fabrics, materials, jewelry, clothing, insects, dust 1/12/2002 WEBS Retreat 5 Application Pull • Dense instrumentation in space and time – environmental monitoring and management – life sciences revolution – emergency analysis and response – surveillance and security • In situ monitoring and management – condition-based maintenance • Ubiquitous computing environments – infer intent from observed action & context, reactive environment • Robotic swarms 1/12/2002 WEBS Retreat 6 Bridging the Technology-Application Gap • Power-aware, communication-centric node architecture • Tiny Operating System for Range of HighlyConstrained Application-specific environments • Network Architecture for vast, self-organized collections • Programming Environments for aggregate applications in a noisy world • Distributed Middleware Services (time, trigger, routing, allocation) • Techniques for Fine-grain distributed control • Demonstration Applications 1/12/2002 WEBS Retreat 7 Where do we stand now? 1/12/2002 WEBS Retreat 8 The de facto platform for sensor nets • Developed a series of wireless sensor devices • • • • • TinyOS concurrency framework Messaging Model Networking stacks (RF and Serial) Multihop routing Several Key components – sensing, logging, data filters, broadcast • Simulation tools 1/12/2002 WEBS Retreat 9 Tiny OS Concepts • Scheduler + Graph of Components init Commands, Event Handlers Frame (storage) Tasks (concurrency) Messaging Component – frame per component, shared stack, no heap • Very lean multithreading • Efficient Layering 1/12/2002 WEBS Retreat internal thread Internal State TX_pack et_done (success RX_pack )et_done (buffer) • Constrained Storage Model init Power(mode) TX_packet(buf) – – – – msg_rec(type, data) msg_sen d_done) • Component: Events send_msg (addr, type, data) Commands power(mode) – constrained two-level scheduling model: threads + events 10 application Application = Graph of Components Route map router sensor appln packet Radio byte bit Radio Packet byte Active Messages RFM 1/12/2002 Serial Packet UART Temp photo ADC SW HW clocks WEBS Retreat Example: ad hoc, multi-hop routing of photo sensor readings 3450 B code 226 B data Graph of cooperating state machines on shared stack 11 Many Research Groups on board • UCB – – – – – – – • • • • • • NEST SensorWeb Blackout Glaser structures CBE BFD BRWC UCLA USC Rutgers winlab Intel Bosch Crossbow 1/12/2002 • • • • • • • • • • • U Wash Rutgers UIUC NCSA U Virginia Ohio State UCSD Dartmouth MIT Accenture and soon many more WEBS Retreat 12 Handful of demonstration applications • 29 Palms • Cory Hall network – ½ million packets over 3 weeks • • • • Surge network and environment display CBE (???) Glaser Shakes Granlibakken retreat watcher => need to get greater application focus • more real and long lived • more dynamics • extract architecture and create framework 1/12/2002 WEBS Retreat 13 Key Experience • Really good at building tinyOS subsystems – non-blocking, split-phase event structures • Internalized the “state of constant change” paradigm – ex: maintain routing tree by constantly rebuilding it – soft state that is always suspect – simple one-way protocols • Operating in the aggregate • Simple mechanisms to accomplish large goals – MAC, ATC • Out of the box on networking abstractions – Low-power listen, wake-up, statistical sampling, weighted aggregation • Understanding of large scale dynamics 1/12/2002 WEBS Retreat 14 Pushing Scale 1/12/2002 WEBS Retreat 15 Good stated agenda with NEST • Develop sequence of open experimental platforms – basic services (time synch, trigger) • • • • • Develop Challenge Application for NEST FSM high-concurrency programming environment Infrastructure support Adversarial simulation Macroprogramming unstructured aggregates Challenge Application composition services coordination synthesis services services SW platform 1/12/2002 WEBS Retreat HW platform 16 New Ideas Secure Language-Based Adaptive • Small, flexible, low-cost, low-power, Service Platform for wireless embedded sensor devices Large-scale Embedded Sensor Networks • Tiny event-driven, robust, open component OS for NEST devices David Culler, Eric Brewer, David Wagner Shakar Sastry, Kris Pister UC Berkeley - mcast, AM, prune algorithmic primitives • FSM high-concurrency prog. env. • Resilient aggregation - for security and other noise • Macroprogramming unstructured aggregates • Adversarial Simulation Impact • Enable creation of embedded distributed syst. of Schedule chal. app defn unprecedented scale and role - 1,000s of tiny networked sensors • Enable new classes of applications integrated with physical world - Greatly simplify creation of distributed systems at extreme scale (HW & SW) - fine-grained distributed control • Accelerate prototyping and evaluation of new OEP1 defn FSM OEP1 on OEP1 eval June 02 lang based optimize final & viz macro. prog. lang env design June 03 June 04 June 05 End June 01 Start OEP2 coord. & synthesis algorithms proto • Enable new, robust basis for distributed, embedded software thru platform design & OEP1 novel tools for simulation and visualizationWEBS Retreat 10x100 kits 1/12/2002 • Drive NW sensor challenge applications log & trace adv. sim OEP2 platform design OEP2 analysis OEP2 OEP3 OEP3 platform chal app & design17 evaluation Wealth of Research Challenges – imperfect operation and reliability service – operating in aggregate network • Create a suite of platforms to initiate the cycle of refinement • New family of issues across all the layers 1/12/2002 WEBS Retreat system architecture technology 18 prog / data model application algorithm / theory – able to be casually deployed in infrastructure (existing or to be created) mgmt / diag / debug • Large numbers of highly constrained (energy & capability), connected devices Technology Challenges • • • • • • • Ultra low power data paths, ADC, ... Radio (CMOS passives, SW, UWB) Optical communication Sensors & Actuators near Zero-power ‘listen’ Basis for ranging & localization Power harvesting and storage – point of sustainable operation • Integration into everyday devices 1/12/2002 WEBS Retreat 19 Node Architecture Challenges • General purpose, application specific optimization • Balance and functionality thresholds • Physical vs virtual parallelism – controller, interconnect hierarchy for network embedded dev. • • • • Passive vigilance Self diagnosis and watch-dog Subsystem abstractions Packaging, packaging, packaging... 1/12/2002 WEBS Retreat 20 Operating System Challenges • • • • • • • • • • Robust, efficient concurrency in situ, dynamic code communication primitives & capabilities scheduling discovery and configuration security programming model synthesis and optimization simulation and testing Security & Privacy 1/12/2002 WEBS Retreat 21 Networking Challenges • Incorporating place and time • Channel management – coding, MAC, hopping • Connectivity Management & Topology formation • Routing – richness, subprimitives, redundancy • Protocols • Architecture – storage, retransmission, naming/addressing – hierarchy • Multicast / Aggregate operation • Scheduling • Security 1/12/2002 WEBS Retreat 22 Service Challenges • Discovery – who are you, what do you have, what do you do, where am I? • • • • • • • • Recruitment of resources Time synchronization Localization Routing services Storage Services Dynamic allocation of resources Negotiation of roles Service Architecture 1/12/2002 WEBS Retreat 23 Distributed Control Challenges • Closed-loop at many levels • network in the feedback loop – observation path – control path • Shift from large centralized analysis to many small points of processing • Distributed constraint solving • Phase-transition characteristics 1/12/2002 WEBS Retreat 24 Programming Challenges • Analysis and synthesis of robust event-driven structures • Optimization for power, jitter, delay bounds • In network programming and code management • Nodal programming model • Programming unstructured aggregates 1/12/2002 WEBS Retreat 25 Application Challenges • It’s all about data – infrastructure data vs in-network data – deep connections between queries and content-based routing – compression vs robustness • Data models for truly long-running queries – when you don’t have much storage and BW, ... • Cooperative processing – multimodal and multi-lateration • Everything has error terms • What are the right higher-level abstractions • Privacy 1/12/2002 WEBS Retreat 26 Algorithms • Classic distributed alg’s face tiny nodes and highly dynamic network structure – power constrained • • • • • • Localization Scheduling Fine-Grain Inverse problems Imaging Constructive foundations of self-organization Understanding how an extreme system is behaving and what is its envelope – adversarial simulation 1/12/2002 WEBS Retreat 27 Overview of Agenda (monday) • Introductions (Culler chair) – Overview of the Project (David Culler) – Closing the loop - Control and Applications (Shankar Sastry) – NEST MICA Platform architecture (Jason Hill) • Dinner • Demo presentations (Brewer chair) – Building-wide monitoring of environment and things (Robert Szewczyk, Anind Dey) – Cheap robots & Smart Dust (Mike Scott & Sarah Bergbreiter) – ROBOMOTE (Gaurav Sukhatme, USC) • Play 1/12/2002 WEBS Retreat 28 Overview of Agenda (Tuesday) • Sensor Processing and Simulation (Sastry chair) – NEST Sensor platform (Alec Woo) – Localization (Kamin Whitehouse) – Visualization and Simulation (Phil Levis) • Breakout Session - Breakthough Opportunities and Challenges – Radical new applications – Key Algorithmic / Theoretical Problems – Novel networking / Systems Design • 12:00 - 4:00 Lunch Recreation and Team Building • Networking and Applications (Pister Chair) – – – – Net dynamics (Deepak Ganesan) Mote geocast (Joe Polastre& Rachael Rubin) Tracking (Bruno & Luca) Security, crypto, beaconing (Adrian Perrig) - key distr, auth • Dinner • Panel: Applications of Networked Embedded Systems Technology (Wagner chair) – Falk Herman, Bosch – Lakshman, Intel - Ubiquitous Office Environments – Kris Pister, Structures, Fire, and all that 1/12/2002 WEBS Retreat want – Alan Mainwaring, Intel - What the field biologists 29 Overview of Agenda (Wednesday) • • • • 8:30 - 9:30 Open Mic 9:30 - 10:00 Break to check out 10:00 - 11:00 Report from Breakouts 11:00 12:00 Visitor feedback 1/12/2002 WEBS Retreat 30