FARM: A Feedback based Adaptive Resource Management for Autonomous Hot-Spot Convergence SystemProblem OverviewTodayTomorrowToday and TomorrowSensor WebSensor Web - SchematicSystem ComponentsSystem Components (contd..)System ModelApplication ScenarioApplication Scenario (contd..)Sensor Web - ChallengesChallenges in building a Sensor Web (contd..)Issues in Autonomous Hot-Spot ConvergenceIssues in Autonomous Hot-Spot Convergence (contd..)Workload CharacteristicsWorkload Characteristics (contd..)FARM: A New Resource Management MethodologyPath-based SchedulingFunctional ModulesFunctional Modules (contd..)Path-level DiagramPathsValue-based SchedulingClosed Loop SchedulingFARM: ArchitectureFARM: Architecture (contd..)Satellite Coordination and Load BalancingGlobal Policies for Distributed Real-Time SystemSurvivabilitySurvivability (contd..)Slide 33SecurityFARM: SummaryFARM: A Feedback based Adaptive Resource Managementfor Autonomous Hot-Spot Convergence SystemS. Swaminathan & G. ManimaranDept. of Electrical & Computer EngineeringIowa State University{swamis,gmani}@iastate.edu{swamis,gmani}@iastate.eduProblem OverviewNASA – Earth Science EnterpriseTo understand Earth system Employs Earth Observing Satellites to collect data about atmosphere, oceans, continentsInformation used to solve scientific mysteries, society problemsOne such problem: HotspotWhat? – Any abnormal natural/man made eventE.g., Volcano, Tornado or Nuclear explosionProblem: Hot-spot detection and locationSolutions called for Collaborative Problem Collaborative Problem SolvingSolvingTodayToday:Today:Large space-based ObservatoriesTomorrowSensor WebSensor Web- Web of satellites with on-board computing powerToday and TomorrowNASA - transition from large observatories to light space-based instruments Instruments with on-board computing and communication powerOn-board computing - must enable hot spot detection, location, and monitoringEnable autonomous problem solvingInstruments placed at different orbits for different levels of accuracyWork as a web of satellites – Sensor WebSensor WebSensor WebSensor WebConstellation of Earth observing satellites Coordination for distributed monitoring, processing, and decision makingCan be considered as an internet of satellitesEnd user must be able to interact with sensor web to get information about hotspotsEasy deployment of technology and scalabilitySensor Web - SchematicSystem ComponentsSatellites:Orbit around the earth at different orbitslow earth orbit (small satellites), geostationary (geostationary satellites), and L1 and L2 orbits (sentinel satellites). Spot beam (coverage area) Data accuracy – depends on orbit level Smaller the spot beam, higher the accuracySystem Components (contd..)Satellite - sensors and instruments for measurementSatellite on-board data processing capabilitieson-board communication capabilitiesCapable of reacting to changes in the spot beamEarth Control Center: Controls/monitors the sensor webEnd user: Queries the sensor webSystem ModelApplication ScenarioScenario - handling of a hot-spotAuto detection or External triggerOn-board processors analyze datacommunicate directly or through other satellitesdifferent data rates for different informationAutonomous satellites detect hot-spotsEstablish contact with control centerIf external triggered - instruments adjustedApplication Scenario (contd..)Scientist becomes aware of hotspotknowledge of location - not requiredTriggers sensor web (nearby satellite) Perform collaborative problem solvingspecific satellite queries relevant satellites detect hotspot and information to control centercontrol center can adjust other satellites (if reqd.)Control center/end user can query any satellite about a hotspotSensor Web - ChallengesInformation TechnologyResource management mechanism - handling uncertain workloadAlgorithms and techniques - data aggregation, compression and image processing.Reconfigurable architectures and algorithms for various on-board data processing.Energy-efficient architectures and algorithms for on-board computing and communication.Satellite TechnologyDynamic control and reconfiguration of satellite instruments.Satellite computing, communication, and instrumentation technologies.Global real-time onboard navigation capability for earth science remote sensing.Challenges in building a Sensor Web (contd..)Domain Specific Technology e.g., water-level monitoring algorithm in Polar Ice caps cloud contamination detection with atmospheric correctionIssue Addressed - Resource ManagementResource ManagementManagement of computing and communication resourcesReliability, availability, performance and security requirementsDynamic re-configuration of sensor webIssues in Autonomous Hot-Spot ConvergenceAutonomous Hot-Spot Convergence Hotspot IdentificationIdentification of satellite(s) to cover the hot-spot for required qualityAllocation of resources - computing and communicationReconfiguration of instrumentation resourcesReliable communication with Earth Control Center/ End User.Issues in Autonomous Hot-Spot Convergence (contd..)Requirements drive the need forLoad balancing among the satellitesDistributed Coordination Schemes for minimizing redundancy in the coverage area Quality-aware distributed scheduling, as the quality of image perceived by one satellite might be higher than anotherFurther, there is a need for continuous monitoring of topology and resourcesTopology Monitoring - knowledge of spot beamResource Monitoring – for resource allocationWorkload CharacteristicsComputational WorkloadStaticDynamicE.g., Default Spot beam coveragePeriodicAperiodicFixed hotspot monitoring e.g, vulcanodynamic hotspot monitoring e.g., tornadoPeriodic computation and communication Continuous handoffReconfiguration of sensorsWorkload Characteristics (contd..)Communication WorkloadInter-satellite - periodic and aperiodicsatellite -> control center - periodic and aperiodicsatellite -> end-user - aperiodic Uncertainty - Computational and Uncertainty - Computational and Communication Communication Workload !!Workload !! Need for Adaptive Resource Management!!Need for Adaptive Resource Management!!FARM: A New Resource Management MethodologyFARM - Feedback-based Adaptive Resource