Demonstration Evening ServiceWave 2010, FIA and FIRE
Maurer, Sakellariou, Brandic : Simulating Autonomic SLA Enactment in Clouds using Case Based Reasoning
1. Simulating Autonomic SLA Enactment in Clouds using Case Based Reasoning Michael Maurer1, Rizos Sakellariou2, Ivona Brandic1 1Distributed Systems Group Institute of Information Systems, Vienna University of Technology Austria maurer@infosys.tuwien.ac.at 2University of Manchester, School of Computing Science, UK
2. Background …. STSM COST Action to UofManchester visiting Prof. RizosSakellariou 3 weeks during March 2010 Presentation in Lyon in June 2010 at focus group meeting for wired networks Results will be published at ServiceWave 2010 in Ghent, Belgium 2
3. Cloud Computing Source: “Buyya, Yeo,Venugopal, Broberg, Brandic. Cloud Computing and Emerging IT Platforms: Vision, Hype and Reality for Delivering Computing as 5th Utility, Elsevier Science 2009.” Automatically adapt to users needs! Challenge: Attaining SLAs vs. optimizing resource allocation Software failures ..... Hardware failures Loadchanges 3
6. 5 Motivation – FoSII Infrastructure Foundation of Self-governing ICT Infrastructures Models and concepts for autonomic SLA management and enforcement Comprises different components MAPE-K cycle LoM2HiS framework Enactor component Knowledge management SLA mapping management … 5
7. 6 Motivation – FoSII Infrastructure Foundation of Self-governing ICT Infrastructures Models and concepts for autonomic SLA management and enforcement Comprises different components MAPE-K cycle LoM2HiS framework Enactor component Knowledge management SLA mapping management … 6
8. SLA knowledge management Simulation Goal of the simulation: Evaluate the quality of a knowledge base in respect to analyzing measurements Input: Measurements (Monitored Metrics) Output: Action to execute Evaluation: Compare the number of SLA violations to the utilization of resources violate as few parameters as possible while utilizing as few resources as possible increase energy efficiency
9. Research Problem Create simulation engine to find suitable knowledge management (KM) system for VM resource allocation Determine interaction of KM with other MAPE-phases Goal of KM: reduce SLA violations allocate resources efficiently basis for increasing energy efficiency 8
12. Simulation Design 11 Plan I: Maps action onto Physical Machines Quality of recommended actions (decisions) = Violations vs provided resources Knowledge base: Recommends action Plan II: Prevents oscillations and schedules execution of actions Analysis I: Queries knowledge base (1) What do we provide? (2) What does the customer utilize? Monitor (simulated): New measurement of an SLA Executor (simulated): Executes action (3) What did we agree in the SLA?
15. CaseBasedReasoning Actions Increase/Decrease storage bandwidth memory parameters to betuned on VM by 10%, 20%, ... Do nothing Future actions: migrate VM outsourceapplication 14
16. Actions Increase/Decrease storage bandwidth memory parameters to betuned on VM by 10%, 20%, ... Do nothing Future actions: migrate VM outsourceapplication 15
21. Conclusion Knowledge management technique-agnostic simulation engine traverses MAPE cycle simulates Monitoring and Execution part evaluates decision making for VM resource allocation Implementation and Evaluation of CBR suitable KM technique reduces SLA violations increase resource utilization leverage learning techniques fine-tune similarity function 20
22. Future Work Evaluation of other KM techniqueswithsimulationengine Rule-basedapproach Default logic Translation of resourceutilization to energyefficiency VM deployment PM management 21
23. Questions & Contact information Michael Maurer Distributed Systems Group Institute of Information Systems Vienna University of Technology Austria email: maurer@infosys.tuwien.ac.at http://www.infosys.tuwien.ac.at/staff/maurer/ 22
24. Outlook Cloud Markets Cooperation with JoernAltmann UofSeoul, South Korea FoSII Infrastructure Planning Execution Actuator Service 1 Knowledge DBs Cooperation R. Sakelariou, Manchester Knowledge Infrastructure Resources ….. Sensor RT Analysis Monitoring Service n Run-time Cooperation with Sztaki, Hungary SLA Manager(s) Sensor Host Host Cooperation with Raj BuyyaUoMelbourne, R. Calheiros PUCRS Resource mapping SLA Violation Propagation, Service Virtualization Monitoring & Metrics Mapping Job Management Interface Input Sensor Values Output Actuator Values Self-management Interface Control Loop Negotiation Interface Knowledge Access
25. Knowledge DBs Predict SLA violations before they happen Problems: How to identify possible SLA violations ahead of time Thresholds for the SLA parameter values where we have to react Tradeoff: preventions of SLA violations vs. doing nothing and paying penalties Consider non SLA parameters like energy efficiency, carbon footprint Possible Solutions: Rules Systems, Default Logic, Situation Calculus, Case Based Reasoning,…
26. Future Work Translation of resource utilization to energy efficiency Development and evaluation of different knowledge management techniques Development of heuristics to selects the most appropriate KM technique VM deployment PM management 25