A Distributed Self-management Service

31
Introduction Architecture Policy language Evaluation Conclusions A Distributed Self-management Service Ioanna Tsalouchidou Universitat Polit` ecnica De Catalunya Kungliska Tekniska H¨ ogskolan [email protected] Supervisor: Leandro Navarro Moldes July 1, 2013 1 / 31
  • date post

    22-Oct-2014
  • Category

    Technology

  • view

    341
  • download

    0

description

Ioanna-Tsalouchidou_Master-thesis-presentation

Transcript of A Distributed Self-management Service

Page 1: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

A Distributed Self-management Service

Ioanna Tsalouchidou

Universitat Politecnica De CatalunyaKungliska Tekniska Hogskolan

[email protected]

Supervisor: Leandro Navarro Moldes

July 1, 2013

1 / 31

Page 2: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

Overview

1. Introduction

2. Architecture

3. Policy language

4. Evaluation

5. Conclusions

2 / 31

Page 3: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

Autonomic ComputingThe CONFINE TestbedMotivation

Overview

1. Introduction1.1 Autonomic Computing1.2 The CONFINE Testbed1.3 Motivation

3 / 31

Page 4: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

Autonomic ComputingThe CONFINE TestbedMotivation

Autonomic Computing

Self-configuration: install/configure according to policies.

Self-optimization: parameter tuning to perform optimally.

Self-healing: identify, trace, diagnose and repair the causefailure.

Self-protection: anticipate, detect and auto-defend againstproblems

4 / 31

Page 5: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

Autonomic ComputingThe CONFINE TestbedMotivation

Overview of the testbed

CONFINE: Community Network Testbed for the FutureInternet

Community networks

Scale up thousands of nodes, distributed geographically

Users allocate resources to run experiments

Multiple experiments run concurrently, sharing resources

5 / 31

Page 6: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

Autonomic ComputingThe CONFINE TestbedMotivation

Architecture

6 / 31

Page 7: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

Autonomic ComputingThe CONFINE TestbedMotivation

Motivation

Address possible challenges in CONFINE testbed

Predefined policies and automatic actions

Self-management system for CONFINE testbed

Policy specification language

7 / 31

Page 8: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

Autonomic ComputingThe CONFINE TestbedMotivation

Related Work

Self management: MyOps, Puppet

Policy languages: KAoS, Rei, Ponder

8 / 31

Page 9: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The Monitoring SystemThe Self-management system

Overview

2. Architecture2.1 The Monitoring System2.2 The Self-management system

9 / 31

Page 10: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The Monitoring SystemThe Self-management system

The monitoring system

Large scale of infrequently-used data

Slice and sliver specific information

Metrics provided by the operating system or synthesized

Receiver-pull model

10 / 31

Page 11: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The Monitoring SystemThe Self-management system

Architecture of the monitoring system

11 / 31

Page 12: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The Monitoring SystemThe Self-management system

The Self-management system

Self-protection

Cooperating with the monitoring system

Components on server and research devices

Collect, policy and action components

12 / 31

Page 13: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The Monitoring SystemThe Self-management system

Architecture of the self-management system

13 / 31

Page 14: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

TheorySyntax

Overview

3. Policy language3.1 Theory3.2 Syntax

14 / 31

Page 15: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

TheorySyntax

Objectives of the language

Simple, extensible, expressive, easy to process

Everything allowed unless explicitly prohibited

If-then-else logic

Describe the stable state of the testbed nodes

Define actions when the system exceeds the stable state

15 / 31

Page 16: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

TheorySyntax

The syntax of the language

Define the resource of reference

CPU, memory, network

One general block for each resource

Smaller blocks within the general for each policy

16 / 31

Page 17: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

TheorySyntax

Example of the syntax I

17 / 31

Page 18: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

TheorySyntax

Example of the syntax II

18 / 31

Page 19: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The MethodScenario CPU

Overview

4. Evaluation4.1 The Method4.2 Scenario CPU

19 / 31

Page 20: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The MethodScenario CPU

Method

Policy scenarios and experiments

No performance metrics

Quality of the design

20 / 31

Page 21: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The MethodScenario CPU

Method

Scenarios for CPU

Definition of the stable state

Actions

21 / 31

Page 22: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The MethodScenario CPU

Scenario CPU I

22 / 31

Page 23: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The MethodScenario CPU

Scenario CPU II

23 / 31

Page 24: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

The MethodScenario CPU

Scenario CPU II

24 / 31

Page 25: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

DiscussionFuture WorkConclusion

Overview

5. Conclusions5.1 Discussion5.2 Future Work5.3 Conclusion

25 / 31

Page 26: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

DiscussionFuture WorkConclusion

Discussion: Self-management system

Self-management system

Cooperation with the monitoring system

Architecture: server and node side

Sliver and slice view

26 / 31

Page 27: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

DiscussionFuture WorkConclusion

Discussion: Policy language

Policy specification language

Easy to comprehend, expressive and extensible

Define resource, stable state, exceeding duration, action

27 / 31

Page 28: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

DiscussionFuture WorkConclusion

Future Work

Aggregation of data during duration period

More fine-grained actions

Prioritizing and grouping

28 / 31

Page 29: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

DiscussionFuture WorkConclusion

Conclusion

Autonomic computing

CONFINE testbed

Self-management system for CONFINE testbed

Policy Specification language

29 / 31

Page 30: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

DiscussionFuture WorkConclusion

References I

Confine project.

http://confine-project.eu/.

T. Koch, C. Krell, and B. Kramer.

Policy definition language for automated management of distributed systems.In Systems Management, 1996., Proceedings of Second IEEE International Workshop on, pages 55–64,1996.

Planetlab.

http://www.planet-lab.org/.

Stephen Soltesz, Marc Fiuczynski, and Larry Peterson.

Myops: A monitoring and management framework for planetlab deployments.

Gianluca Tonti, Jeffrey M. Bradshaw, Renia Jeffers, Rebecca Montanari, Niranjan Suri, and Andrzej Uszok.

Semantic web languages for policy representation and reasoning: A comparison of kaos, rei, and ponder.In International Semantic Web Conference, pages 419–437, 2003.

30 / 31

Page 31: A Distributed Self-management Service

IntroductionArchitecture

Policy languageEvaluation

Conclusions

DiscussionFuture WorkConclusion

A Distributed Self-management Service

Ioanna Tsalouchidou

Universitat Politecnica De CatalunyaKungliska Tekniska Hogskolan

[email protected]

Supervisor: Leandro Navarro Moldes

July 1, 2013

31 / 31