Digital On-Demand Computing Organism for Real-time Systems Dod Org SPP OC Kolloquium DFG SPP 1183 “Organic Computing” Nürnberg, September 15/16, 2011 KIT – The cooperation of Forschungszentrum Karlsruhe GmbH and Universität Karlsruhe (TH)
Talk Overview Motivation and Overview Current Work Phase III: Organic Hardware Organic Monitoring Organic Low Power Management Organic Middleware DodOrg Demonstrator Platform Interaction and Overview Scenarios and Results Conclusion 2 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
DodOrg Motivation Classic Scenario: DodOrg Scenario: Only those scenarios can be handled System reaction based on indications That were considered in advance, (higher level of abstraction) Where the cause can be detected, E.g. CRC/bit error rate, network bottleneck, environmental change or change on Where the corresponding reaction had application level been explicitly programmed. Proper reaction possible even if Lack of adaptation leads to insufficient reactions (e.g. shutdown …) Scenario was not considered in advance Cause was not detected Reaction was not explicitly programmed Flexible response to changed environmental situation Scenario detection: recognize that something is different Adapt to changed requirements either by Demonstrator platform known path or gradual process of rearrangement (optimization, healing) 3 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
DodOrg: Refined Layer Model Brain Level Self-Adaptation Application Application Testbed Nervous Self-Optimization Monitoring System (all groups) Self-Healing Biological Considerations (Brändle) Application API Organic Monitoring System (Karl) Proactive Intelligent Data Analysis Middleware Monitoring, Organic Middleware Stable Hormone Organ Level Feedback Interaction (Brinkschulte) Heart Hormone Level Computation Temperature, Thermal-aware Distributed Low Power Local Traffic Energy Management (Henkel) distribution Dynamic Power Management Cell Level Real-time considerations Myo- cardial Hardware OPC Extension Organic Processing Cell Monitoring Cells (Becker) Stability Aspects 4 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Phase III: Project Objectives Stability Robustness The ability of the system to Extending the stable system provide the required service property towards more serious while reacting upon external system changes . and internal events. + Fault resistance + Oscillation avoidance + Increased tolerance + Normal operating conditions - Increased overhead - Faulty components 5 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Organic Hardware Approach (Prof. Becker) Modularity Heterogeneous Array of Organic Processing Cells (OPCs) Same blueprint for all OPCs OPC with common Common infrastructure structure but with FPGA DSP I/O specific functionality Cell Cell Cell Cells easily replaceable Local intelligence Memory Monitor FPGA artNoC Router N Configuration Cache Cell Cell Cell E artNoC Power-Management artNoc S • broadcast Router Monitoring Configuration Low Level Monitoring W • real time Channel • adaptive routing Allocation/Release Configuration-Management State Interface FPFA µ Proc I/O Cell-Specific Configuration Cell Cell Cell Messenger L Functionality Control Interfaces (μProc, DSP, Power Status FPGA,FPFA, Network Interface Power Control Monitoring Monitor Status Memory, Peripheral Observer Control Devices Clk local Monitoring, Middleware Clk global etc.) Observer Clock and Power Low Power Management Monitoring Data Emergency Calls Management (DVFS) Flexibility Cell Data path Reconfigurable data path 6 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Organic Processing Cells: Robustness (Prof. Becker) Robustness during development phase OPC-Lifecycle Scope Loading new configuration Establish-inter-cell-data path Development Phase: Power up/down cells Reconfiguration Method: On-demand hardware monitoring Ongoing Change Blank configuration pattern Robustness during processing phase Scope Processing Phase: OPC-data path (packet sender) OPC to OPC communication path Calculation (artNoC-Network) Goal: hardware support for cell immune system Method: artNoC header packet protection Channel auto release 7 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
On-Demand Hardware Monitoring (Prof. Becker) T3 T1 T2 T2 Vertical Routing Oscillator Horizontal Macro Routing T4 • Memory requirement (T1-T6, OSC) : 302 Bytes T5 Routing Base • 80µs/ routing connection • 880µs/ OSC macro • Reconfiguration time: 1.5ms (8 bit ICAP) T6 T1 • Suitable to monitor large areas 8 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Organic Processing Cells: Robustness (Prof. Becker) Challenge: Fault introduced communication deadlock OPC-2-OPC communication: Wormhole Flow Control Flexible Low buffer requirements Packet spread across several routers Failure/attack affects several routers Protect control flits: Corrupted header-flit misrouting of packets Checksum Missing tail/header-flit blocked virtual channels If downstream cell – release channel via feedback line If upstream cell – inject tail flit to release channel with error code 9 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Synthesis Result: ArtNoC Router on Xilinx FPGA (Prof. Becker) Slices RT: Real-Time FB: Feedback Channel 4000 3500 WF: West-First Routing 3000 BC: Broadcast 2500 FA: Full adaptive Routing 2000 RC: Packet Recovery 1500 PR: Control-Flit Protection 1000 2VCs 500 4VCs 3VCs 0 3VCs 4VCs 2VCs ca. 110 additional Slices for Control-Flit Protection Avoidance of blocked Virtual Channels (VCs) Avoidance of misrouting packets 10 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Organic Monitoring (Prof. Karl) Objectives Application Application Testbed Self-Adaptation Monitoring (all groups) Coordinated, cooperative and Self-Optimization Biological Considerations (Brändle) Application API Organic Monitoring System (Karl) Proactive Intelligent Data Analysis system-wide monitoring Self-Healing Middleware Monitoring, Organic Middleware Feedback Stable Hormone Fundamental for self-organizing (Brinkschulte) Interaction Hormone Level Computation systems Temperature, Distributed Low Power Local Traffic Stable Energy Application Hardware Providing monitored data to Management (Henkel) Distribution Dynamic Power Management Organic Middleware and Thermal Real-time considerations Hardware Requirements Organic Processing Cells Management for further analysis Monitoring OPC Interaction, (Becker) Status Status Metrics Providing self-awareness Monitoring Self-Awareness Prerequisite for all self-X features Raw & Cooked Configuration Feedback Ability of system state determination Data Ability of system state classification Permitting the comparison of two arbitrary system states Middleware Thermal Management 11 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Organic Monitoring: State Evaluation (Prof. Karl) Using a rule-based approach Properties Learning evaluation rules in a Rules convert an occurrence ratio of an dedicated training phase event into a fitness value Defining a normal system state One rule per event / hormone Comparison of further system states Weighted arithmetic mean for determ- with the normal state ining the fitness value for the system Different states then can be compared by comparing the fitness values Monitored Value Monitored Value 12 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Organic Monitoring: State Classification (Prof. Karl) Using k-means clustering for definition of individual system state at runtime Treating all available event occurrences as a point in a n-dim space Clustering of close points within this space Using euclidean distance for online state detection 13 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Organic Monitoring: Phase Detection and Prediction (Prof. Karl) Goals: Prediction of future system states A Identification and avoiding of potentially harmful system states B State Prediction: Using a runlength-encoded Markov chain as C predictive model Trained in a dedicated learning phase using the previously classified system states D Current Predicted Past System States Past System States Past System States System State System States 14 9/16/2011 SPP 1183 OC Kolloquium – Nürnberg, 15.-16. September 2011
Recommend
More recommend
