180 likes | 189 Views
This project aims to develop a network virtual machine that abstracts and guarantees real-time behavior for unreliable and mobile networks, enabling coordination and control in sensor, actuator, and processor networks.
E N D
NEST System Working GroupMeeting #1 Jack Stankovic University of Virginia 11-13 September 2001 Boeing Huntington Beach, CA
Presentation Outline • Project Overview • Research Products • Anticipated OEP Integration and Collaboration
Project Overview Title: A Network Virtual Machine for Real-Time Coordination Services Jack Stankovic, PI University of Virginia Partners: (UIUC, CMU, LM)
Goal • Create a network virtual machine that is a coordination and control layer (middleware) that • abstracts (API) • controls, and • guarantees aggregate real-time behavior for unreliable and mobile networks of sensors, actuators, and processors.
Sensor/Actuator Clouds Resource management, team formation: real-time, mobility, power Heterogeneous Sensors/Actuators/CPUs (macro motes) • battlefield awareness • pursuer/evader
Middleware Components Applications Real-time Service APIs In-network processing Data placement Node Placement Congestion control Routing Schedulability analysis RT MAC Run-time Service Local RT CPU Scheduling
Research Questions • Invention of new lightweight protocols that can support guaranteed aggregate behavior • real-time • power • mobility • limited processing and memory capacity • large scale • Solutions based on • diffusion type algorithms with aggregation • randomness • feedback control principles/RT/ uncertainty/overload control • MMDP • real-time scheduling theory
The Team REAL-TIME SERVICES Lockheed Martin Virginia Aggregate Control FC/ RT Applications Req. Team Coord. Mobility/ Wireless Data Discovery MMDP/ Power CMU Illinois RT/ Power
1. Technology Development Program • Berkeley OEP: • Real-time run-time services with aggregate performance guarantees (UVA and UIUC) • Power management and control in wireless communication (UIUC and CMU) • Merge results • Coordinating visit between UVA and UIUC in August • Coordinating visit between UIUC and LM • Boeing OEP • Consensus protocols (UVA and Xerox Parc): coordinating visit already completed in August
2. Product Type X___Operating system software X___Middleware _____Application software (noise control for Boeing OEP, pursuer/evader for Berkeley OEP) _____Configuration software X___Algorithms / theoretical foundations (please describe within comments) _____Tools (please describe) _____Other (please describe)
Technology Areas Coordination Services Time-Bounded Synthesis Service Composition and Adaptation Embedded Software Focus Areas Middleware Services Configurations On- line Off- line Fill In Taxonomy For Project Technology Challenges Res. Mgmt CPU Memory Network Power Fault Tolerance Time Synchronization Heterogeneous Processing Safety Criticality
4. Challenge Areas 4. Challenge Area Classification:(Indicate all challenge area(s) targeted by your research.) a. Lifecycle: Is your technology targeted to: ____ design time (e.g. tools, techniques used during system development) X___ run time (e.g. online software) b. Domain: Is your technology targeted to: _____ application domain (e.g. noise control, pursuer evader games) X____ solution domain (software/system design related issues, e.g. middleware) c. Solution domain issues: Is your technology targeted to: _____ fault detection X____ online reconfiguration (possibly in response to faults) _____ offline configuration _____ time synchronization X____ group membership (online determination of group members) X____ group consensus (collaboration of group members towards aggregate goals) X____ probabilistic methods _____ other (please describe)
5. Collaborations • a. OEP collaboration: Are you expecting at this point to work with: • X____ Berkeley OEP • X____ Boeing OEP • b. Group I collaboration: • Xerox Parc: Consensus protocols for Boeing OEP • Intra-group (UVA, UIUC, CMU, and LM): Real-time services for Berkeley OEP
6. Integration Interface a. Provided interfaces Berkley OEP: high level APIs specifying environmental queries, commands and their aggregate real-time requirements (periods and deadlines) EXAMPLES: activity(area,event,ST,DU,P,D,MP,e) set_lifetime(L) b. Required interfaces: Boeing OEP: noise model and local vibration control Berkley OEP: Fine except for more memory, perhaps offload more from CPU to HW
7. OEP Framework Requirements • X___ Network (e.g. wireless, TTA, Ethernet, Fibre-channel), specifically: • Berkeley OEP: wireless • Boeing OEP: Switched Ethernet • ____Operating system, specifically: • X___Threads (e.g. single-threaded, preemptive multi-threaded), specifically: • Berkley OEP: preemptive multithreaded architecture • X___Scheduling protocols (e.g. static/dynamic, RMS, EDF, MUF), specifically: • A real-time scheduling policy, e.g., EDF or RMS • X___Other (describe): • Berkeley OEP • More RAM/ROM for code and data • Additional HW handling bits from radio
Scalability and Training 8. Scalability a. Number of nodes: what network sizes are targeted by your technology: O(103) b. Memory: what node memory requirements are targeted by your technology: >100KB 9. Training Requirements: Real-time scheduling theory Control theory Markov decision process Our APIs - compliant with the TinyOS component interface
Releasibility Restrictions • No proprietary claims. • No releasibility restrictions.
Schedule 1/02 Motes 3/02 RT-MAC 8/02 RT Directed Diffusion 10/02 Data Placement/Node Placement 12/02 Schedulability Analysis 12/02 Integrate Power Management