Software Methodologies for distributed systems Policy-based - - PowerPoint PPT Presentation

A Fuzzy, Utility-based Approach for Proactive Policy-based Management

Software Methodologies

for distributed systems

Christoph Frenzel, Henning Sanneck, and Bernhard Bauer RuleML 2013, July 11 – 13, Seattle, WA, USA

Software Methodologies

for distributed systems

Policy-based Management (PBM)

The system should make complex decisions guided by operational objectives, Challenge: increase the level of automation Systems Management with Policies by operational objectives, e.g., maximize capacity. The system should act proactively in order to avoid problems

Software Methodologies

for distributed systems

Rate eliability Problem Threshold

Classical PBM and Its Problems

Call Setup Success R Re

Software Methodologies

for distributed systems

Rate eliability Problem Threshold

Classical PBM and Its Problems

Sharp distinction between acceptable and unacceptable system states

• Reactive behavior

Call Setup Success R Re

Policy conflicts are resolved with complex rules that interweave technical knowledge and

• perational objectives
• Costly maintenance
• Reactive behavior

Software Methodologies

for distributed systems

Utility-based Rule System

Concept of Proactive PBM

Fuzzy Logic System

Replace boolean predicates with continuous memberships to allow reasoning in inaccurate domains ECA rule-based Policy system (technical knowledge) with utility-based conflict resolution (business objectives) Fuzzification of monitoring events to create fuzzy events indicating their severity Inference of the value of actions based on fuzzy rules weighted with utilities Defuzzification by selecting actions according to their value

Software Methodologies

for distributed systems

Utility-based Rule System

Concept of Proactive PBM

Fuzzy Logic System

Replace boolean predicates with continuous memberships to allow reasoning in inaccurate domains ECA rule-based Policy system (technical knowledge) with utility-based conflict resolution (business objectives) The action value represents the degree of rationality and considers: Fuzzification of monitoring events to create fuzzy events indicating their severity Inference of the value of actions based on fuzzy rules weighted with utilities Defuzzification by selecting actions according to their value degree of rationality and considers:

• 1. Events the action handles
• 2. Utility of the events‘ treatment
• 3. Severity of the events

Software Methodologies

for distributed systems

Rate eliability Problem Threshold

Behavior of Proactive PBM

Call Setup Success R Re

Software Methodologies

for distributed systems

Rate eliability Problem Threshold

Behavior of Proactive PBM

Replacing sharp thresholds with fuzzy jeopardy zone

• Proactive behavior based on fuzzy

event levels

Call Setup Success R Re

event levels Policy conflicts resolved by comparing the action values

• Separation of technical knowledge

and operational objectives

Software Methodologies

for distributed systems

System Design

Fuzzificationof monitoring events to create fuzzy events Inference of the value of actions based on fuzzy rules weighted with utilities Defuzzificationby selecting actions according to their value

Software Methodologies

for distributed systems

Event level determination

› Annotate event with fuzzy event level › 3 KPI states: acceptable, unacceptable, and jeopardy › Memberships can be computed by any function provided as an event

Fuzzification

R e l i a b i l i t y P r

• b

l e m E v e n t L e v e l

any function provided as an event specification

Software Methodologies

for distributed systems

Value computation

› Fuzzy rules are technical knowledge:

IF reliability problem IS raised AND ret available IS true THEN action IS ret optimization WITH objective_dcr

Fuzzy Inference

Event Condition Action Utility of Operator Objective WITH objective_dcr

› Objectives are defined using utilities

Utility of Operator Objective

Software Methodologies

for distributed systems

Value computation

› Combine expected utilities of rules to overall value

» Domain-dependent aggregation, e.g., sum » Maximum of the rules for one objective to avoid double counting

Fuzzy Inference

Software Methodologies

for distributed systems

Conflict resolution

› Resolve action conflicts by selecting actions with higher value

» Constraint optimization problem

Defuzzification

Action Value Action

selected

Software Methodologies

for distributed systems

Scenario taken from mobile networks management

› Problem situations and objectives created at random

Fuzzy, utility-based PBMS has best performance

› 15% better than fuzzy PBMS

Evaluation

0,241 0,273 0,315

0,1 0,2 0,3 0,4

Classical PBMS Fuzzy PBMS Fuzzy, utility-based PBMS

Average Value

› 15% better than fuzzy PBMS › 31% better than classical PBMS

Classical PBMS Fuzzy PBMS Fuzzy, utility-based PBMS Fuzzy events No Yes Yes Utilities No No Yes

Software Methodologies

for distributed systems

The presented approach models …

› … a Utility-based Policy System with › … a Fuzzy Logic System.

Thus, the system enables …

› … automatic control of the system guided by operational objectives encoded as utilities and

Conclusion

encoded as utilities and › … proactive actions triggered by fuzzy event levels.

In the future, we are working on…

› … include observations, e.g., from ineffective actions › … modeling approach for the operator objectives & technical knowledge › … include stochastic actions and estimate their effectiveness using machine learning

Questions?

Software Methodologies

for distributed systems

christoph.frenzel@ds-lab.org