Delft Cooperation on Intelligent Systems SMDS, a top-down approach to Self-Management for Dynamic Collaboration Systems

Transcription

1 SMDS, a top-down approach to Self-Management for Dynamic Collaboration Systems Bernard van Veelen bernard.vanveelen@decis.nl

2 Presentation Overview Context SMDS Overview SMDS Approach esults Future Work 2

3 Context Dynamic Collaboration System (DCS): Multiple platforms with various resources joining in the pursuit of a common mission Composition may vary over course of time Environment Mission 3

4 Context Bottom up approach to DCS, determine plan, discover peer, contract peers, (who have to do the same cycle) Start execution Will not: Fit Critical environments (time, safety, mission) Global Optimization, runtime improvement, Crossover flexibility (make use of new resources, repair when resources leave, re-use processing), Convergence, Escalation-detection 4

5 Context Dynamic Collaboration System Goals Infer, Plan Management Sub-system Coordinate System of Systems Act World Monitor Observe Users Needs Triggers on Needs Constraints set by Goals (Mission) Provide integrated Solutions and coordination Monitor and adjust Management Sub-system is System of Systems 5

6 Use Knowledge Base to document System s capabilities SMDS Overview Use Automated easoning technology to infer collaboration blueprints (solutions) in response to emerging needs Maximize performance Guarantee safety, security and other QoS requirements Compile combined solutions, respecting each requirement Track progress/quality of aspects of combined solution 6

7 SMDS Overview Segments Planning Monitoring Instantiation emember: Knowledge based Delegate Management esponsibilities Provide Open but Total, Integrated Solutions Quality of Management Managed Sub-systems 7

8 SMDS Aproach: Planning Segment Context egisters Needs, Mission (Goals) Disseminates global plan (MasterPlan) Planning Monitoring Instantiation Purpose Provide global plan Match supply/demand Safeguard Mission/Goals 8

9 Planning Segment: Functional Decomposition MasterPlan : egister Needs and Mission/Goals (global constraints) Compile Solutions for Needs into MasterPlan easoning : Determine Solutions for Needs Knowledge Base Capture knowledge regarding domain, capabilities, status. Mission, Goals Needs MasterPlan easoning Knowledge Base Planning Monitoring Instantiation Domain Capabilities Status MasterPlan QoS-Info 9

10 SMDS Approach: Instantiation Segment Context Implements MasterPlan Planning Instantiates/Configures Sub-systems Monitoring Instantiation Purpose Implement Global Plan Start/stop required services Send signals/messages to sub-systems Configure infrastructure Process transport Load-management 10

11 Instantiation Segment: Functional Decomposition LoadManagement Divide labour among hosts Planning Instantiation Present on each host Maintain bookkeeping Monitoring Instantiation Formulate messages, incantations, etc Plumbing Low-level start- stop and configuration services Instantiation Plumbing New Process MasterPlan Signals System Map Load Mgmt QoS info Messages 11

12 SMDS Approach: Monitoring Segment Context Samples Status of sub-systems Planning Evaluates Progress of MasterPlan Monitoring Instantiation Provides QoS information Purpose Guard Quality of Monitor System Health Trigger improvement actions: Instantiation Segment Planning Segment 12

13 Monitoring Segment: Functional Decomposition QoS Management Evaluate info from Monitoring/QoS eports Decide whether improvement required Evaluate Progress of MasterPlan Feedback eceive QoS eports from (intelligent) subsystems Monitoring (on each host) Sample hardware, communications and software Planning Monitoring Instantiation Monitoring Feedback QoS eport MasterPlan QoS info QoS Mgmt 13

14 SMDS Approach: Segment Context eceives contact-information of sub-systems Provides System Topology Map Planning Monitoring Instantiation Purpose Federate Systems/s Map System topology map Notify relevant s to federate Provide Functionality (lib) 14

15 Segment: Functional Decomposition Exchange contact info with servers on other platforms Planning Federate services Issue federation notifications Provide System Topology Map Used by Instantiation segment Monitoring Instantiation Library XYZ-Server A XYZ-Server B System Map message Contact info 15

16 Benefits Global Overview Optimization QoS management Escalation triggers Flexibility Federated Set-up obustness Detect and repair (QoS management) Efficiency Batch allocations Combine (parts of) Solutions 16

17 esults Prototype: COMPASS Configuration, Organization and Management Prototype for Autonomous Systems of Systems Coordination Challenge at DECIS Lab first results. 17

18 Future Work Coordination Challenge Obtain more results for more complex scenarios ICIS-A work QoS management, DataGrid s, Security Management Hybrid Approaches, Layering, etc 18