Pg 1 of xx AGI www agi com Communications & Data Dave Finkleman

Pg 2 of 40 AGI www agi com Overview Introduction to Mobile ad-hoc Networking –OSI Layers –Routing –Disruptive Phenomena Model Selection –Physical Layer –Link, Network, and Transport Layers Missile Defense communication example Conclusion

Pg 3 of 40 AGI www agi com Mobile ad-hoc networking ( MANET ) Self-organizing networks of –Dynamically mobile elements –With equal technical ability –Independent of fixed infrastructure or centralized control General characteristics –Dynamic and often unpredictable network tolopogy –Variable capacity, often congested, bandwidth limited links –Energy and power constrained –Low physical security

Pg 4 of 40 AGI www agi com Desirable capabilities Scalable –From few widely distributed elements through multitudes of dense, randomly mobile elements – is not scalable (10 nodes, 300 meters) –Bluetooth is neither scalable nor true MANET Master-slave relationship 100 meter range

Pg 5 of 40 AGI www agi com Desirable capabilities Mobile elements with static infrastructure (cellular, WAN, LAN) and fixed infrastructure networks (Internet) Spectrum of route discovery and route maintenance schemes Data link layer operation with a variety of embedded network and transport layer protocols

Pg 6 of 40 AGI www agi com OSI framework: MANET Open Systems Interconnect (OSI) Physical layer (the environment) –Environment must be able to support communications –Channel loss (obstructions, atmospherics) –Interference –Mechanical and physical incompatibilities Data link layer (the road map) –Must establish pathways within the communications medium –Make physical links consistent (MAC)

Pg 7 of 40 AGI www agi com OSI framework: MANET Network layer (the route) –Must accommodate changes in topology and discover routes (IP) Transport layer (the traveler following the route) –Must match delay and dropout characteristics to sustain reliable communication Application layer –Must be able to handle frequent and unanticipated disconnections (HLA)

Pg 8 of 40 AGI www agi com Routing focus Minimizing routing overhead Control packets consume bandwidth Minimizing delays –Links break and make randomly, interrupting ongoing communication Optimizing paths –Many approaches sacrifice route optimization in order to diminish overhead and delays High altitude platforms can address all of these concerns

Pg 9 of 40 AGI www agi com Routing focus Preventing loops –Keeping route discovery from looping back on itself Minimizing computational effort and storage –Route discovery and maintenance require complex logic and significant router memory Scaling –Bandwidth and overhead can be totally consumed by internal route discovery and maintenance –Especially if every node maintains real time routing tables for every other node High altitude platforms can address all of these concerns

Pg 10 of 40 AGI www agi com Routing issues Location-dependent carrier sensing –Carrier sensing performed at transmitter - most influenced by phenomena near receiver –Hidden terminals: Node out of range of sender but within range of receiver Communication between the receiver & hidden node burdens channel –Exposed terminals: Node within range of sender / out of range of receiver Transmissions from exposed node deceive sender –Channel is occupied

Pg 11 of 40 AGI www agi com Routing issues Poor collision detection Incoming signals weaker than transmissions Acknowledgements (or lack of), latent –Potentially leading to unnecessary retransmission

Pg 12 of 40 AGI www agi com Hidden terminal

Pg 13 of 40 AGI www agi com Exposed terminal

Pg 14 of 40 AGI www agi com Exposed terminal

Pg 15 of 40 AGI www agi com Mitigating routing issues Link layer protocol approaches –Proactive: regularly scheduled route discovery –Reactive: route discovery in response to link loss Architectural approaches –Flat: all nodes equal –Hierarchical: some nodes elected or designated for special functionality Hybrid approaches –Some proactive, some reactive –Some flat, some hierarchical

Pg 16 of 40 AGI www agi com Fixed & ad-hoc routing Fixed networks: exploit static routing tables –Distance vector: “shortest” path –Link state: avoid contention and broken links Bellman-Ford: In any graph there exists a spanning tree, a set of arcs that visits every node exactly once

Pg 17 of 40 AGI www agi com Fixed & ad-hoc routing ad-hoc network analogies –Distance vector ad-hoc on demand distance vector (AODV) (reactive/flat) Dynamic source routing (DSR) (reactive/flat) –Link state Optimized link state routing (OLSR) (proactive/flat) Ad-hoc networks require a richer routing scheme

Pg 18 of 40 AGI www agi com High altitude platform advantages Ability to reach widely distributed inter-cluster nodes No hidden or exposed nodes –Potentially all hybrid inter-cluster nodes visible Ability to assist geographically aided intra-cluster routing –GPS enabling such as cell phones use, for example, potentially half duplex may be sufficient

Pg 19 of 40 AGI www agi com High-altitude platform advantages Availability of sufficient power and processing –Accommodates techniques whose overhead or energy requirements would swamp most mobile nodes MANET characteristics are not compromised, since The platform is not acting as a central controller

Pg 20 of 40 AGI www agi com Measures of performance Network size (number of nodes) Network density Network capacity Connectivity structure (number of neighbors) Mobility pattern (speed, range, …) Link bandwidth Traffic pattern (packet size, type of traffic, …) Link characteristics (bidirectional, unidirectional) Transmission medium (single vs multi-channel)

Pg 21 of 40 AGI www agi com Modeling and simulation OPNET-STK Example Simulation configuration –Clusters Land Vehicles with TBRPF and WRP (few, modest mobility) Aircraft Cluster with AODV and DSR (few, high mobility –Inter-Cluster Enabled by HAA Realizations with ZRP and DREAM for comparison Tradeoffs –Efficiency (hop count) and other measures of performance –Compare combinations of two element alternatives above

Pg 22 of 40 AGI www agi com M&S selection criteria 1.Represent at least four lower OSI layers (1-4) 2.Represent discrete events 3.Represent highly non-linear complex systems Capable of broad statistical analysis and Monte Carlo 4.Represent wireless interface from emission through propagation to reception Encompasses antenna placement & characteristics, refraction, diffraction, absorption, & scattering

Pg 23 of 40 AGI www agi com M&S selection criteria 5.Represent mission environment including mobility characteristics of potential nodes & other matters affecting data exchange 6.Represent diversity of land, sea, air, near-space, space nodes & platforms 7.Generate or accept data traffic from missile defense mission & other subscribers

Pg 24 of 40 AGI www agi com M&S selection criteria 8.Flexibility and Scalability Expansion to arbitrary numbers of distributed, nodes open and modular software design 9.Ease of use for the purpose intended 10.Cost-modeling & simulation must fit the resources allocated to project

Pg 25 of 40 AGI www agi com Major modeling candidates Network Simulation/Parallel-Distributed Network Sim: Discrete event simulator targeted at networking research (public domain) Dynamic Network Emulation Backplane Project: Brings multiple network simulators together in single experiment GloMoSim/Parsec: Scaleable, discrete event, network simulation. Library for the Parallel Simulation Environment for Complex Systems (PARSEC) parallel, discrete event, simulation language (public domain)

Pg 26 of 40 AGI www agi com Major modeling candidates QualNet: Derived from GloMoSim. Well-supported & maintained COTS product SSF (Scalable Simulation Framework): Includes SSF Network Models (SSFNet), with "open-source Java models of protocols, network elements, & assorted support classes for realistic multi-protocol, multi-domain Internet modeling & simulationSSF Network Models (SSFNet) Dartmouth SSF (DaSSF): Process-oriented, conservatively synchronized parallel simulator

Pg 27 of 40 AGI www agi com Major modeling candidates OMNet++: Component-based, simulation package suitable for traffic modeling, protocol analysis & evaluating complex software systems. OPNET: Leading commercial network simulator, including "library of detailed protocol and application models. Widely used to diagnose performance of real-world networks & adjust or reconfigure them. MLDesigner (MLD): Integrated platform for modeling & analyzing architecture, function & performance of high- level system designs

Pg 28 of 40 AGI www agi com Physical layer modeling Real world events require more than descriptions of interconnects Models & simulations of physics & dynamics of executing missions evolve actions that enable “event driven” network models & simulations Only two enterprises support the needs of this effort well: –FreeFlyer, produced by AI Solutions, COTS suite that supports the entire specific mission operations lifecycles –Satellite ToolKit (STK), produced by Analytical Graphics, Inc STK can evaluate complex in-view relationships among dynamic space, air, land and sea objects instantiated in great physical detail

Pg 29 of 40 AGI www agi com Co-simulation, hardware in the loop, advanced waveforms STK-OPNET Co-simulation STK-OPNET Object exchange OPNET, OPNET Modeler STK,STK-COMM STK-V0, STAMP (MFT) Analysis approach Dynamic Comm Paths, Latency, LOS, Doppler Histories Aggregated latencies & network performance Dynamic Comm Paths, Latency, LOS, Doppler Histories Application of Existing Mobile Ad-Hoc Networking Techniques New or advanced routing, network discovery, protocols & latency- tolerant techniques

Pg 30 of 40 AGI www agi com ICO CommSat Representative mission geometry, sensors & links Full Duplex Comms with Interceptors in Flight High Altitude Airship

Pg 31 of 40 AGI www agi com Initial Defensive Operations (IDO)

Pg 32 of 40 AGI www agi com Active links with interceptor

Pg 33 of 40 AGI www agi com Raw accesses to interceptor

Pg 34 of 40 AGI www agi com Doppler histories & free space losses Approx. 30 db less path loss than to lowest accessible commsat More favorable physical geometry than to geo comsats More manageable two-way doppler variations

Pg 35 of 40 AGI www agi com OPNET analysis Modulation scheme driven by Doppler dynamics Network discovery & routing driven by node mobility & nomadicity Protocols driven by required latency, reliability & security Implementation driven by device envelope, power generation & thermal management, mechanics or electronics of beam formation & steering

Pg 36 of 40 AGI www agi com OPNET: Interceptor communications “Wired” links Protocol tailoring Node processing & buffering Error correction schemes Network performance capability vs. realized

Pg 37 of 40 AGI www agi com OPNET: Interceptor communications

Pg 38 of 40 AGI www agi com Missile defense comms

Pg 39 of 40 AGI www agi com Conclusion STK facilitates analysis of modern mobile ad-hoc networking –“Mobile Networking with Strong Physical Layer Interactions” Physical phenomena occur on time scales comparable to, or less than, network transactions –Hypervelocity vehicles (including satellites) –Interplanetary missions STK matched with “wired” network simulations –Event-driven –Physical-world “frozen” during network transactions (OPNET) Efficiently and uniquely suited for wireless, mobile, ad-hoc networks –We have demonstrated significance of intimate interactions among lower-four OSI layers

Pg 40 of 40 AGI www agi com Summary Introduction to mobile, ad-hoc, networking –OSI layers –Routing –Disruptive phenomena Model selection –Physical layer –Link, network & transport layers Missile defense communication example Conclusion

Pg 41 of xx AGI www agi com BACKUPS

Pg 42 of 40 AGI www agi com Flat routing comparisons N : # of nodes in the network e : # of communication pairs in the network

Pg 43 of 40 AGI www agi com Hierarchical routing comparison N : # of nodes in the network e : # of communication pairs in the network M : average # of nodes in a cluster H : # of logical levels L : average # of nodes in a logical groupG : # of logical groups in the network

Pg 44 of 40 AGI www agi com Location-assisted routing comparisons N = # of nodes in the network e = # of communication pairs in the network M = Average # of nodes in a cluster