policy based autonomic decision making for wireless networks

24
Policy Based Autonomic Decision Making for Wireless Networks Niki Gazoni, Fontas Fafoutis {ngazoni,fontas}@csd.uoc.gr Department of Informatics and Telecommunications University of Athens, 2007

Upload: lane-walters

Post on 30-Dec-2015

24 views

Category:

Documents


1 download

DESCRIPTION

Policy Based Autonomic Decision Making for Wireless Networks. Niki Gazoni, Fontas Fafoutis {ngazoni,fontas}@csd.uoc.gr. Department of Informatics and Telecommunications University of Athens, 2007. E2RII. End-To-End Reconfigurability Phase II http://e2r2.motlabs.com - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Policy Based Autonomic Decision Making for Wireless Networks

Policy Based Autonomic Decision Making for Wireless

Networks

Niki Gazoni, Fontas Fafoutis {ngazoni,fontas}@csd.uoc.gr

Department of Informatics and TelecommunicationsUniversity of Athens, 2007

Page 2: Policy Based Autonomic Decision Making for Wireless Networks

E2RII

• End-To-End Reconfigurability Phase II– http://e2r2.motlabs.com

• Intelligent Wireless Communication System– Main Characteristics

• RAT Interoperability• Context Aware• Autonomic • Reconfigurable

Page 3: Policy Based Autonomic Decision Making for Wireless Networks

RAT Interoperability

• Multiple Radio Access Technologies– WLAN, WiFi, WiMAX– GSM/GPRS, UMTS, HSDPA– Bluetooth, Infrared

Page 4: Policy Based Autonomic Decision Making for Wireless Networks

Context Aware

• Aware of their environment– Location– Device Capabilities

• Static (Hardware, OS, Java VM)• Dynamic (Memory, Battery, CPU, Signal Strength)

– User Preferences • Tariff Class, Network Preferences

– Network Characteristics• Static (Capacity, Coverage)• Dynamic (Congestion, Delay, Available Bandwidth)

– Service Requirements• Resources

– etc

Page 5: Policy Based Autonomic Decision Making for Wireless Networks

Autonomic

• Minimize human interference – Self – Management– Self – Configuration– Self – Awareness– Self – Healing– Self - *

Page 6: Policy Based Autonomic Decision Making for Wireless Networks

Reconfigurable

• Dynamic Adaptation based on Contextual Information– Protocol Adaptation (Transport Layer)

• A protocol reconfigures its congestion control algorithm to a more suitable according to the monitored traffic

– Service Adaptation (Application Layer)• Video conference application “downgrades” to phone

conference when network cannot support video requirements due to heavy load

• Seamless– Transparent to the User

Page 7: Policy Based Autonomic Decision Making for Wireless Networks

Basic Architecture

Service Provisioning

Self-Configuration / Self-Management

Autonomic Decision Making

Context Management

Page 8: Policy Based Autonomic Decision Making for Wireless Networks

Autonomic Decision Making

• Wireless Communication World – Modeled in OWL

• Policies define the behavior of the System– Expressed in SWRL rules

• Decision Making Process– Contextual Information is imported into the Ontology– Reasoner infers knowledge based on rules and

context– Decision is parsed from the output of the reasoner

Page 9: Policy Based Autonomic Decision Making for Wireless Networks

Protégé

• Developed by Stanford University– http://protege.stanford.edu

• Graphical Ontology Designer– Developed in Java– Open Source– Supports OWL / RDF Ontologies– Supports SWRL Rules bound to the Ontology– Supports Plug-ins (Reasoners)– Provides Java Library: protégé.jar– Differences between Protégé OWL implementation and standard

OWL • Unique Name Assumption (UNA)

– Survey (2007) suggests that 70% of academic researchers use Protégé as ontology editor

Page 10: Policy Based Autonomic Decision Making for Wireless Networks

Ontology - OWL Classes• Device

– Cell phone, Laptop, PDA• Network

– Instance that has specific operator and uses specific RAT

• Service– Classified based on requirements– SMS, Phone Call, File Transfer,

Web, Video Streaming• Area

– User Location, Coverage• CostProfile

– Free, Economic, Advanced• RAT

– GPRS, UMTS, WiFi

Page 11: Policy Based Autonomic Decision Making for Wireless Networks

Ontology – OWL Properties

• Values– Provided by the Context Management module– Produced by the Reasoner

• Example Properties– hasCost (Network → CostProfile)– hasService (Network → Service)– PrefersOperator (User → Operator)– hasLowBudget (User → boolean)– isCustomer (User → Operator

Page 12: Policy Based Autonomic Decision Making for Wireless Networks
Page 13: Policy Based Autonomic Decision Making for Wireless Networks

Expressing policies in SWRL rules

• Policy continuum• Different stakeholders, different views

– Business view, system view, administrator view, device view ...

• Goals– Choosing the most appropriate network– Ability for handovers– Automated procedures (e.g mass updates)

• Restrictions while using SWRL

Page 14: Policy Based Autonomic Decision Making for Wireless Networks

Choosing the most appropriate network

• isAccessible– Set of networks that are available in the area and the

user can connect to• isProposed

– Subset of available networks that support the desired service

• isPrefered-Cost– Subset of previous networks that comply with user’s

budget profile• isPrefered

– Subset of previous networks that comply with user’s preferred provider profile

Page 15: Policy Based Autonomic Decision Making for Wireless Networks

Example Rule

• a.PayedNetworks

isInArea(?d, ?a) ^

availableAt(?n, ?a) ^

hasType(?n, ?r) ^

supportsRAT(?d, ?r) ^

operatedBy(?n, ?o) ^

usesDevice(?u, ?d) ^

isCustomer(?u, ?o) →

isAccessible(?d, ?n)

• Device needs to be in network’s coverage

• Device must support network’s RAT

• User must be customer of network’s operator

• Device can connect to the network

Page 16: Policy Based Autonomic Decision Making for Wireless Networks

Jess Rule Engine

• Reasoner– Developed in Java– Free for Educational Use

• http://herzberg.ca.sandia.gov/jess

– Can be imported in Protégé – Implements Rete Algorithm

• Charles Forgy Ph.D. Thesis

• How it works– Translates ontology knowledge into Jess Facts– Reasons Jess Facts using Rete Algorithm– Translates new Jess Facts back into ontology knowledge

• Jess Facts– (assert (property_name domain_class range))

Page 17: Policy Based Autonomic Decision Making for Wireless Networks

Example Scenario• Context

– Two Operators • UoA• RWNO

– Three Networks • UoA-Wifi (Free)• RWNO-GPRS (Economic)• RWNO-WiFi (Advanced)

– A User• UoA Student• Low Budget• Preferred Operator: UoA

• Scenario– User is roaming while consuming

a low requirement service• Objective

– Decide on the network which best fits to the contextual information

RWNO-GPRS

RWNO-WiFi

UoA-WiFi

Page 18: Policy Based Autonomic Decision Making for Wireless Networks

Network Selection• (assert (isAccessible MobilePhone

RWNO-GPRS))

• (assert (isProposed MobilePhone RWNO-GPRS))

• (assert (isPrefered-Cost MobilePhone RWNO-GPRS))

RWNO-GPRS

RWNO-WiFi

UoA-WiFi

• (assert (isAccessible MobilePhone RWNO-GPRS))

• (assert (isAccessible MobilePhone RWNO-WiFi))

• (assert (isProposed MobilePhone RWNO-WiFi))

• (assert (isProposed MobilePhone RWNO-GPRS))

• (assert (isPrefered-Cost MobilePhone RWNO-GPRS))

• (assert (isAccessible MobilePhone RWNO-GPRS))

• (assert (isAccessible MobilePhone RWNO-WLAN))

• (assert (isAccessible MobilePhone UoA-WiFi))

• (assert (isProposed MobilePhone RWNO-WLAN))

• (assert (isProposed MobilePhone RWNO-GPRS))

• (assert (isProposed MobilePhone UoA-WiFi))

• (assert (isPrefered-Cost MobilePhone RWNO-GPRS))

• (assert (isPrefered-Cost MobilePhone UoA-WiFi))

• (assert (isPrefered MobilePhone UoA-WiFi))

Page 19: Policy Based Autonomic Decision Making for Wireless Networks

Implementation

• So far everything runs on Protégé Platform

• How to implement it into a Java application?– protege.jar– jess.jar– swrl-jess-bridge.jar

Page 20: Policy Based Autonomic Decision Making for Wireless Networks

Protégé-OWL API

• Parse an ontology file into an OWLModel class– OWLModel Constructor

• ProtegeOWL.createJenaOWLModelFromURI(String uri)

• Update the OWLModel– OWLModel methods

• OWLNamedClass getOWLNamedClass(String name)• OWLProperty getOWLProperty(String name)• OWLIndividual getOWLIndividual(String name)• void createOWLIndividual(String name)• void addPropertyValue (OWLProperty prop, Object value)

Page 21: Policy Based Autonomic Decision Making for Wireless Networks

SWRL-Jess-Bridge API

• Links an OWLModel object to a Rete object (jess.jar)– SWRLJessBridge Constructor

• SWRLJessBridge(OWLModel model, Rete rete)

• Using the bridge we can control the reasoner– SWRLJessBridge Methods

• void resetBridge()• void ImportSWRLRulesAndOWLKnowledge()• void ExportSWRLRulesAndOWLKnowledge()• void RunRuleEngine()• void WriteAssertedIndividualsAndProperties2OWL()• String[] getAssertedProperties()• int getNumberOfAssertedProperties()

Page 22: Policy Based Autonomic Decision Making for Wireless Networks

ADM Module

• Initialization– Import the owl file containing the ontology, the policies

and (relatively) static knowledge into an OWLModel– Create a Rete object and bridge it with the model

• Decision Making in 5 steps– Update the OWLModel with dynamic knowledge– Infer knowledge using the bridge– Clear the OWLModel from useless knowledge– Parse the asserted properties– Return the decision

Page 23: Policy Based Autonomic Decision Making for Wireless Networks

Difficulties

• Measurements on a desktop computer show a slow decision making procedure– Caching– Re-code some modules in C++

• Impossible to run it on cell phones– Libraries in use do not run in J2ME– Much slower decision making due to limited

computing power

Page 24: Policy Based Autonomic Decision Making for Wireless Networks

Thank You

Questions?