Linda Briesemeister, José Joaquin Garcia-Lunes- Aceves, Hamid R. Sadjadpour, José Meseguer, Mark-Oliver Stehr, Carolyn Talcott DARPA PI Meeting August 2, 2005 Washington DC Robust Internetworking in Dis r r u u p p tive Environments

RIDE Team SRI International Computer Science Laboratory Carolyn Talcott (PI), Mark-Oliver Stehr (co-PI), Linda Briesemeister Formal Modeling and Analysis Automated Reasoning Tools Semantic Models Policy Languages Secure Sensor networks Resource Management and Scheduling Dependable System Architecture University of California, Santa Cruz Baskin School of Engineering José Joaquin Garcia-Luna-Aceves (PI) Hamid R. Sadjadpour Wireless, mobile Ad-Hoc networks Fault-Tolerant Internetworking Multipoint communication Communications and Signal Processing University of Illinois at Urbana-Champaign Department of Computer Science José Meseguer (PI) Mark-Oliver Stehr (co-PI) Software Architectures Secure, multicast protocol analysis Formal Interoperability Reflection and meta-programming Petri Nets August 1st

RIDE/SPINDLE Embedding Forwarding Control Convergence Layer Application FSKBBS Knowledge Management Route Planning Bundle Forwarding Resource Management Knowledge Base Bundles/FramesLink StatusKnowledge Bundle Store Frame Store Trigger

RIDE Algorithms Opportunistic Message Switching –Takes into account content and resistance Coordinated Resource Scheduling –Maintain a virtual traffic infrastructure Reflective route planning –Generate robust routing plans Distributed Information Management –Uniformly manage routing-related information

Opportunistic Message Switching Accomplishments –Analyzed mobility-capacity-delay tradeoffs of wireless networks in which nodes store and carry packets before delivery to destinations R. de Moraes, H. Sadjadpour, and J.J. Garcia-Luna-Aceves, ``Mobility-Capacity-Delay Trade-off in Wireless Ad Hoc Networks,'' Ad Hoc Networks Journal, to appear. –Developed the Space-Content-adaptive-Time Routing (SCaTR) framework, which enables data delivery in the face of temporary and long-lived MANET partitions. –Implemented SCaTR by extending AODV: Proxy takes custody of bundle if no direct route available, e.g. due to disconnection/disruption –Showed that the performance of SCaTR is better than on- demand and epidemic routing using simulations in GloMoSim with scenarios of varying degrees of random or predicted connectivity.

Coordinated Resource Scheduling Accomplishments –Special-purpose Java simulator inspired by DTNRG simulator allows random and predefined dynamic topologies –Design and Java prototype of a baseline algorithm for resource management based on a virtual infrastructure –Extension of network simulation model with a physical resource schedule based on annotated Petri nets

Reflective Route Planning Accomplishments –Preliminary Definition of Knowledge Representation and Routing Plans –Baseline Planning Algorithm based on Symbolic Search in Maude –Interoperation between Java Simulator and Maude Planner ? 1 2

Milestones and Costs End of August 05 –Design of Models and Information Exchange Algorithms –Preliminary prototype in Java: Distributed information management End of April 06 –Final designs and prototypes in Java/Maude: Distributed information management Reflective route planning Coordinated resource scheduling –Final design and prototype in GloMoSim: Opportunistic message switching Funds Status: –Spending slightly below expectations due to delays with subcontracts –Transition of Mark-Oliver Stehr to SRI will enhance collaboration and SRI subcontract will be increased correspondingly –No changes in total cost and timeline Preliminary report after 6 months Final reports after 14 months Technical reports:

Go/No Go Criteria In scenarios with 20% availability and 80% utilization bundles will be delivered eventually (i.e. 100% reliability) assuming buffers are sufficiently large and network is sufficiently connected Tradeoff between reliability and storage space will be improved by distributed mechanisms to discard superfluous bundles In scenarios with local congestion, resource management will reduce congestion and improve delivery rate In scenarios with predictable behavior, route planning will improve delivery rate compared with a base line shortest path routing algorithm

Highlights after Phase I Reduction of congestion due to active resource management Show superiority of opportunistic message switching approach compared to traditional routing and epidemic approaches. Use of a formal planning engine as a core component for reflective routing –integrates multiple routing algorithms and –increases robustness of routing

Towards Phase II Expected Accomplishments after Phase I –New resource-driven paradigm for routing –Unique combination of opportunistic and formal planning- based routing Remaining Problems after Phase I –For a single algorithm: Parameter selection and adaptation –For multiple algorithms: Algorithm selection, collaboration, overall interoperation –Lack of Solutions for Security and Trust –Lack of integration between MAC and network layers –Limited support for multipoint communication with different degrees of end-to-end reliability.

Towards Phase II Three-pronged approach to address remaining problems –Leverage UIUC/SRI’s expertise on formal methods and AI technology -> build upon ongoing work at SRI to integrate these two –Leverage UIUC/SRI’s expertise on formal approaches to network security -> propose new approach to address security and trust issues - Leverage UCSC/SRI‘s expertise on routing algorithms and multipoint communication -> propose algorithms for multipoint communication and enhance cross layer integration

Preliminary Results

RIDE: Hello Window ● These values are averaged over a longer period (Hello Window). ● Node maintains the past n values, and averages them to produce its current contact value.

RIDE: Simulations ● SCaTR framework is added to AODV in GloMoSim. ● Several Mobility Scenarios: ● Compared to flooding, controlled flooding, AODV, AODV with source buffering.

RIDE: Delivery Rate (Varied Scenario Size)

RIDE: Delivery Rate (Varied Scenario Length)

RIDE: Control Overhead (Varied Scenario Size)