1 Integrity Service Excellence Complex Information Systems 19 Mar 13 Robert J. Bonneau, Ph.D. AFOSR/RTC

2 Goals: Preserve critical information structure and minimize latency over a heterogeneous distributed network and system Ensure network and system robustness and stability under a diverse set of resource constraints and manage not assuming static models Find invariant properties for a given network and system from a distributed set of observations and predict network behavior Develop unifying mathematical approach to discovering fundamental principles of networks and system and use them in network and system design Payoffs: Preserve information structures in a network rather than just delivering packets or bits Quantify likelihood of a given network management policy to support critical mission functions Predict and manage network and system failure comprehensively Complex Networks and Systems

3 Foundations of Information Systems - Model heterogeneous distributed systems using unified mathematical framework through previous measurement and validate - Verify the properties of a given system application through measurement of a limited set of system parameters and assess mission risk - Define general architectural principles of design through unified assessment of system operating properties - Generalize design properties to universal system architectural principles Program Objectives Payoff - Assess and verify properties of a distributed heterogeneous system where there is limited access to its elements - Assess dynamic Air Force system mission performance and assess risk of failure

4 System Properties Information Systems Research Measure System Information & Verify Properties Diverse Types of Systems Complex networks and systems uses measured information to assure, manage, predict, and design distributed networks, systems, and architectures Complex Networks and Information Systems Roadmap Dynamic, Heterogeneous, Air Force Systems Critical Information System Measurement Local Network/Systems Research Assure Critical Information Delivery Network/Systems Management Research Manage Information Flow Global Network/Systems Research Predict Network Performance

5 Local Network/System Research: Preserving Information Content Statistical geometric coding structures are used to transport diverse sets of information in a network and system and preserve its critical structure Information Timescale t Content Information Distribution Content Information Loss With Interference Content Information Recovery Less: Latency/Computation/ Storage More: Information Loss With Interference Less: Information Loss With Interference More: Latency/Computation/ Storage Recovered Information Loss Distributed Information Loss Measurable Information Loss Significant Information Source Deterministic/Minimal Coding (ex: Trellis Code) Hybrid Code (ex: Network Code) Random Code (ex: Rateless Code) t packets, variables, registers, Recover Using Coding Recover With Code and Retransmit Recover With Retransmission

6 Less: Information Loss With Disruption More: Latency, Difficult to Control Less: Latency More: Information Loss With Disruption, Controllable Information Sources Information Timescale t Protocol/Policy Information Distribution Protocol/Policy Information Loss With Interference Protocol/Policy Information Recovery Source 1 Source 2 Source 3 t groups of packets, subroutine, virtual mem. The state of information transfer on a network changes with network and system management policy and protocol – Particularly important to the Air Force given its unique heterogeneous mobile infrastructure Network/System Management Research: Guaranteeing Information Transfer Recovered Information Message 1 Message 2 Message 3 Deterministic Routing (ex: OSPF) Hybrid Routing (ex: OLSR) Random Protocol (ex: Flooding) Recover With Redundancy and Retransmit Recover With Redundancy Recover With Retransmission Information Loss Distributed Information Loss Measurable Information Loss Significant

7 Less: Latency/Disruption Tolerant More: Controllable Less: Information Loss Under Disruption More: Latency, Resource Intensive Information Sources Information Timescale t Architecture Information Distribution Architecture Information Loss With Interference Architecture Information Recovery Source 1 Source 2 Source 3 Recovered Information Message 1 Message 2 Message 3 t blocks of information, program, virtual memory We wish to develop information invariants that can be used to assess network/ system performance Global Network/System Research: Architecture Performance Invariants and Prediction Deterministic Routing (ex: Core/Backbone) Hybrid Network (Mesh) Random Network (ex: Mobile Ad Hoc) Reroute Information Reroute and Change Distribution Change Information Distribution Information Loss Distributed Information Loss Measurable Information Loss Significant

8 Example: Unified Mission Assured Architecture Current networks are managed with multiple protocols depending on their taxonomy Air Force networks, particularly Airborne Networks are heterogeneous A unified network approach should adapt to the conditions and provide design principles Less: Disruption Tolerant, Latency More: Information Loss Under Interference, Observable/Controllable Less: Information Loss Under Interference, Observable/Controllable More : Disruption Tolerant, Latency Design Principles According To Constraints Adapt According To Measurements

9 Foundations of Information Systems Hybrid Architecture Hybrid Protocol Hybrid Content Random Architecture Random Protocol Random Content Deterministic Architecture Deterministic Protocol Deterministic Content Measured Performance Regions Heterogeneous Information Network States (packets, packet blocks, packet groups) Software States (variable, subroutine, program) Hardware States (register, ram, virt. mem) System Measurements Less: Information Loss Under Disruption/Live More: Latency, Resource Intensive/Safe Less: Latency/Disruption Tolerant/Safe More: Controllable/Live Best Integrated Performance Region (timescale/level of abstraction ) Global Properties Statistical Properti es Stable/Resourced Secure Unstable/Un-resourced Insecure Measure and verify information system properties among various system constraints

10 Measuring Information Systems Fundamental Properties Units of information translate across heterogeneous domains and can be used to measure and quantify system performance - Taking this approach can lead to a unified systems and security strategy Deterministic Protocol Distribution Time Evolution (Global Properties) Deterministic Heterogeneous Random Content (local) System Policy/ Protocol (management) System Structure (global) Deterministic Content Heterogeneous System Heterogeneous Protocol Deterministic System (1/information timescale) Frequency Data Network Packet Groups Packet Blocks Wireless Network Modulation Unit Waveform Signal Array Hardware/ Software Register/ Variable Ram/ Subroutine Virtual Mem./ Program Social Words Phrases News Reports/ Blogs Biological DNA Protein Synth. Cell Function Basic Information Unit Scales Digital Systems General Systems Random Protocol Random Content Heterogeneous Content Random System Measured System Properties Not Resourced, Not Stable, Not Secure Design Excluded Properties Resourced, Stable, Secure, (Safe) Design Included Properties

11 Algorithms for Information Networks If we would like to estimate, detect, control, or predict networks, there are many algorithms that have been adapted to the relevant network conditions We would like new classes of integrated algorithms that can adapt across many dynamic network conditions Dynamic/ Non-stationary Static/ Stationary Random Deterministic Time Frequency/ Scale Stationary Markov Dec. Process Min-max Estimation Wiener Filter Adaptive Matched Filters Kalman Filter Bootstrap Methods Extended Kalman Filter Sequential Probability Ratio Tests Particle Filtering Architecture Hybrid Statistics Protocol/ Policy Content More: Robust to Change/Computationally Intensive Less: Robust to Change/Computationally Intensive Critical Space of Network Performance Necessary Algorithm Properties

12 Comprehensive Systems Modeling - Model heterogeneous distributed systems using unified, modular, composable and scalable mathematical framework from previous measurement and system specification - Use new statistical, algebraic, and geometric representations and theory for modularized representations and composable into a modeling framework Unified Representation Modular, Composable, Scalable Model of Unified System Resource Policy Security Framework Database Arch. Operating System Prog. Languages Design Tools Mission Applications Physical Environ. Resource Const. Processing Hardware Network System of Interest Mathematical Models Statistical, Algebraic, Geometric, …

13 Measurement-Based System Verification - Verify the properties of a given unified system through measurement of a limited set of parameters and calculate system risk of not meeting mission requirements - Assess risk by distance between properties of desired representation (model) and measured properties - Incorporate risk of sparse measurement Desirable Properties: (Example) Robustness to Disruption Undesirable Properties: (Examples) Latency, Interference, Computational Overhead Measurement Mission Requirements Performance Verification Low Mission Risk Medium Mission Risk High Mission Risk Risk Assessment Measured Properties Desired Properties

14 Measurement Validation Trade-space - Define general application architectural and policy validation principles through unified assessment of system operating risk - Apply to existing architectures through policy implementation Architecturally Validated Modalities (low mission risk) Architecturally Excluded Modalities (high mission risk) System Operating Trade-space

15 Mission Performance Guarantees Cloud Component Space Airborne Terrestrial Introduce Advanced Mathematical and Modeling Techniques Into System Components Advanced Mathematical Algorithm Current & Future System Component System Components Complex Information Systems Current & Future Architectures Introduce measurement algorithms and components into existing systems architectures Use measurement based verification strategies to assure mission performance Statistical invariants for modeling based on measured data to validate models Incorporate algorithms into new generations of semiconductors for distributed online system assessment Systems Components in Architecture + Future Mathematical Systems Analysis