Autonomic Administration - HAL 9000 Meets Gene … · Autonomic Administration: HAL 9000 Meets Gene Roddenberry ... DCN DRS CPNI Sales Comp PB Awards ... Paging CNR PBITS MP/F FLEXCOM

Autonomic Administration - John Strassner Page 1

Autonomic Administration:HAL 9000 Meets Gene Roddenberry

��

��

�� ! "#��! �� "��


Strassner Presentation Styles� ��$

% � ��&��#��&��&

� � ��$% ��'��((��

� #��$% "��(��)��*��

� ! ��$% � �� #� ��

� #��$% +�'�

� ��$% �� ,��(��(-.

� ��$% �� &��/�&�&��,��/.


Motivation


A Traditional OSS/BSSExternal data flow

Internal data flow

Pending data flow

System appears twiceNAME

Planned systems

Key:

JOBMOBE

TMM

NI

DFG

PVI

CID/SAM

LOC

INPLANS

ISIS

ASOS

Mech Eng

TESS

CCP

TOPS

NRSS

Billstar 3

Billstar 1PCDB

EC

RIMSPRECISEPay by

Phone

DOMSCustomizer

800

Custom800 DB

RCRMS

BOSSCOR

OSMOPLIDB

E911

NAA

LMOS

Exch Plus

APTOS

PREMIS

MIStarwriter

CMS(CCRS)

SABR

LFACS

FWS

PICS/DPCR

TNDS/TK

LEIS

WM

COSMOS

SOAC

PAWS

LOMS

FIRST

ORGIS

GIRBAIF

IRSS

SDIDSDDL-POF

SOCS

PDR

TOR

AOG

SORD

CRMS

EARSMAPS

CESAR

CABS

Data Svc

PR

JOUR

GL

RAP

3rd Pty

Bill Print

MP USAGE

RM REVEBill Format

Bill Day

Billing

Directory

Delivery

Corp Books

MI

CL CONF

EM EXCHListing Svc

C/CA

SOFE

LSD&C

Data

WarehseFIMS

PARIS

CARTS

COR

SUMMIT 4.0

SBIR

MRDB

Sales Agency

ORBITSAthena

AA PBCC

SMS

DRSDCN

CPNISales Comp

PB Awards

APTOS

PaSS

COIN

AP

PBVS

CSFT

IP

SPACE

MARCH

CSTAR

Separation

ConnectVu

CLONES

TNM NMA-F

DCOS-2000

NetPilot

SEAS

EADAS NDS-TIDE

AMOSNSDB

TIRKS

IPMS

MOPICS PMM

CMTS

FTDM

SARTS

PVS | PMI

REACT2001

TSA

NTAS

MTAS

LATIS

CRAS

CIASANS

LMOS

MLT

SORD

SOAC

PBOD

Service

Manager

Electronic

Bonding

ALRU

PDS-ERA

PDS

CustomerProfile

AIM

POS-R

POS

BRIS

REMS

ESS

EmFiSys

TAPS

AUTS

TRAINS

WTS

IFS

TWISTATR

NSDM

MTR

TCMS

Tech PDP

PMIS

PagingCNR

PBITS

MP/F

FLEXCOM

COSMOS

WFA/C

OPAS/Loopview

OPS/INE

Predictor

INATransport

SSI PBRIMS IPMSCCSN TANCCPL MP/F

ComnLang Taskmate

SCSFEPS

CUR/CAR

TAGS

FDOC

Network

AT&T

Network

DSCNOR AT&T

Network

NOR

STPSCPISCP

EDIIS

PB1

Network

AT&T

NOR

IS

ERMIS

Common Interface Layer

ATC

CIDB

Advantage

TIRKS


Name: JohnSCUSTID: 12345

Shortcomings - Infrastructural

� ��% *��$

% �$��0��(��1��

% �((��0��

% ��((��2��'��3

� "��&��% "��*��

% �� )��&��

% "��&��4�� 0��&��

% #�� &�$�$�� 5 ��

Name: Strassner.JohnID: “12345”


Current Network Management Deficiencies

� �&&��&�� $��6��1��1��)��(��1�� '�� &�� )��1��)��

� �� $�� 1$�� (��$�� $��((��

� 7��1��$�� $�� 2&��1�$��3(�� (��'�� &��(��

� ��8��8��2�� 9��9��9��(3 1�'9��$�


More Effects – Constituency Separation� *� ��)�� &��

% � ��&�� )��/�)��&��

� ��&�� '��)��1��

% 7�� :� ��7� ��&�

% /�'�� &��

% ��(�� '�� &�

% �� $�� &�


Potential Loss of Revenue to Failure

;<=>�?@A� ��

;B�BAC�?C=��

;DD<�EA?F�'��&

;B�AE?�?@?��

;CE@�CBD��

;B�>==�=<B"� �� #��&$

;B�?A?�==="��

;<>E�A>A+��

;B�=D@�B>=��"��

;B�<BA�<=@� �� &

;?�A<<�?=@#��

;?�EBC�E=<:��&$

��

Source: Disaster Recovery: Reaction, Not Reality, Meta Group, February 2004


So What IS Autonomics?


Machines will take over all management tasks, rendering humans superfluous.

Future Vision of Autonomic Computing?

Hal 9000, 2001

��


Machines will free system administrators to manage system at a higher level

Future Vision of Autonomic Computing

��


Autonomic Networking

� Technical complexity: human body � technology, devices� Business complexity: macro-economics � e- and m-Commerce� Behavioral complexity: social interaction � service composition� Operational complexity: healing � anti-virus, configuration management

Biology, Sociology, and Economics can Inspire Better Networks!

Complexity abounds!

InferencePlane

CognitiveFunctions

InferencePlane

Network Planning

and Optimization

Involuntaryfunctions

Management Plane

ConfigurationManagement

Management Plane

PolicyCont r oller

PolicyBroker

0. . n

0. . 1

0. . n

0. . 1

Br okerCoor dinat es

ScopedPolicyDet ails

DecisionRequest edDet ails

PolicyAct ionPer f orm edDet ails

PolicyDom ain

0. . n0. . 1 0. . n

HasPolicySubDom ains

0. . 1

M anagem ent Dom ain

PolicyApplicat ion

0. . n0. . n

0. . n0. . n

ScopesPolicyApplicat ion

PolicyExecut ionCom ponentPolicyDecis ionC om po

nent

1. . n1 1. . n1

Direct sExecut ionO f

M anagedEnt it y

0. . n 0. . n0. . n 0. . n

ScopedM anagedEnt it ies 0. . n

0. . n

0. . n

0. . n

PolicyAppI nvolvedWit h

0. . n

0. . n

0. . n

0. . n

PolicyAct ionPer f or m edO n

0. . n

0. . n

0. . n

0. . n

PolicyDecisionRequest edBy

PolicyVer if icat ionCom ponent

0. . n

0. . n

0. . n

0. . n

PolicyVer if icat ionPer f or m edO n

PolicyVer if icat ionPer f orm edDet ails

PolicyAdm inist rat ionCom ponentCom ponent

PolicyVer if icat ionCom ponentCom ponent

PolicyProxyCom ponentCom ponent

PolicyBrokerPolicyCont roller Com ponent PolicyCont roller

Applicat ion

PolicyEnf or cem ent Com ponentCom ponent

PolicyApplicat ion

PolicyDecisionCom ponentCom ponent

PolicyExecut ionCom ponentCom ponent

M anagedEnt it y

P olicy St or a geC om ponen tCom ponent

. . . . . .. . . . . .

B rok er Coo rdin at es

. . . . . .. . . . . .

HasPolicySer ver Com ponent s

1

. . .

1

. . .

Dir ect sExecut ionO f

. ..

. . .

. ..

. . .

PolicyAppI nvolvedWit h

. . .

. . .

. . .

. . .

PolicyDecisionRequest edBy

. . .

. . .

. . .

. . .PolicyAct ionPer f orm edO n

Policykeywords : SequenceO f St r ingpolicyNam e : St r ing

PolicyDom ain PolicyApplicat ion

PolicyCondit ionAt . . .PolicyEvent At . . . PolicyCondit ionNo. . . PolicyAct ionAt om icact ionSequenceNum ber : I nt egerhasExecut ed : I nt eger = 0

get SubAct ions( )

PolicyAct ionNonSt dact ionDat a : O ct et St r ingact ionEncoding : St r ingact ionResponse : Boolean

PolicyEvent No. . .

ECAPolicyRuleAt om ichasSubRules : Boolean = FALSEisCNF : Boolean = TRUE

get SubRules( )

PolicyAct ionCom posit eact ionExecut ionSt rat egy : I nt eger = 2act ionSequence : I nt eger = 1act ionsCNF : Boolean = TruehasSubPolicyAct ions : Boolean = FALSE

get Act ions( )

PolicyCondit ionCom po. . .PolicyEvent Com posit eCom posit e

PolicyAct ionact ionO rder : I nt eger

PolicyCondit ioncondit ionI sNegat ed : Boolean = FALSE

PolicyEventevent SequenceNum ber : I nt egerhasEvent Evaluat ed : I nt eger = 0

get Event s( )

ECAPolicyRuleenabled : I nt eger = 1isM andat oryEvaluat ion : Boolean = TRUEpolicyValidFor : Tim ePer iodpr ior it y : I nt eger = 0usage : St r ing

I sECAPolicyRuleWellFor m ed( )

ECAPolicyRuleCom po. . .

PolicySubject

M anagem ent Policy

PolicyTar get

. . .. . . . . .. . .

P olic yCo nt a ine dIn

. . .. . . . . .

HasPolicySubDom ains

. . .. . .. . . . . .. . .

PolicyAppUsesPolicy

. . . . . .. . . . . .

ScopesPolicyApplicat ion

. . .

. . .

. . .{ or der ed ; m us t be >= 1

PolicyAct ionAt om ic}

. . .

HasPolicyA ctio ns

. . .

. . .

. . .

{ or dered; m ust be >= 1PolicyCondit ionAt om ic}

. . .

HasPolicyCondit ions

. . .

. . .

. . .

{or dered; m ust be >= 1PolicyEvent At om ic}

. . .

HasPolicyEvent s

. . . . . .. . . . . .

{ ord er ed;>=1 PolicyAct ionAt om ic | |>=1 PolicyAct ionNonSt d}

ECAPolicyUsesPolicyAct ion

. . .

. . .

. . .

. . .

{or der ed ;>= 1 Polic yCo ndit ion Ato mic | |>= 1 PolicyCondit ionNonSt d}

ECAPolicyRuleUsesPolicyCondit ion

. . .

. . .

. . .

. . .

{ or der ed;>= 1 Polic yEv ent At om ic | |>= 1 PolicyEvent NonSt d}

ECAPolicyRuleUsesPolicyEvent

. . .

. . .

. . .

. . .

{or dered; m ust be >= 1ECAPolicyRuleAt om ic}

HasECAPolicyRules

. . .

. . .

. . .

. . .

Subject I nt er act sWit h

. ..

. . .

. ..

. . .

Target I nt er act sWit h

Ent it y

Resour ceSt at e. . .

ResourceSt at ist ical. . .

Fault I nf o

Conf igur at ion. . .

Account ing. . .

Per f or m ance. . .

Secur it yI nf o

CLI Me th od

RM O NM et hod

SNM PM et hod

TL1M et hod

Propr iet aryM et hod

ServiceSt at e. . .

Ser viceSt at ist ical. . .

Descr ibedM gm t I nf oDet ails

M anagem ent I nf om gm t I nf oValidFor : Tim ePer iodr et r ievalM et hodCurr ent : I nt egerr et r ievalM et hodsSuppor t ed : SequenceO f I nt eger

M anagem ent M et hodEnt it y M anagedEnt it y

Suppor t edM gm t M et hodDet ail

. . .

1

. . .

1M gm t I nf oO bt ainedBy

. . .

1

. . .

1

Descr ibedByM gm t I nf o

. . . 1. . . 1

Suppor t edM gm t M et hods

Ent it y

ResourceSt at e. . .

ResourceSt at ist ical. . .

Fault I nf o

Conf igurat ion. . .

Account ing. . .

Per f orm ance. . .

Secur it yI nf o

C LI M e tho d

RM O NM et hod

SNM PM et hod

TL1M et hod

Pr opr iet ar yM et hod

Ser viceSt at e. . .

ServiceSt at ist ical. . .

Descr ibedM gm t I nf oDet ails

M anagem ent I nf om gm t I nf oValidFor : Tim ePer iodr et r ievalM et hodCur r ent : I nt egerr et r ievalM et hodsSuppor t ed : SequenceO f I nt eger

M anagem ent M et hodEnt it y M anagedEnt it y

Suppor t edM gm t M et hodDet ail

. . .

1

. . .

1M gm t I nf oO bt ainedBy

. . .

1

. . .

1

Descr ibedByM gm t I nf o

. . . 1. . . 1

Suppor t edM gm t M et hods

ATM I nt er f acebit Rat eType : I nt eger

Et her net I nt er f aceduplexSet t ing : I nt egeret hernet Cur rent Speed : I nt egeret hernet Speed : I nt eger

DeviceSubI nt er f ace LoopbackI nt er f aceSer ialI nt er f acebandwidt h : St r ingclockRat e : St r ingisDCE : Boolean

TokenRingI nt er f acer ingO penSt at us : I nt egerr ingSpeed : I nt egerr ingSt at e : I nt egerr ingSt at us : I nt eger

NullI nt er f ace

DeviceI nt er f acecust om er I nt er f aceNum ber : St r ingdeviceConf igM et hod : I nt egerisI nt er f aceEnabled : Booleanvendor I nt er f aceNum ber : St r ing

M ediaI nt er f ace LogicalI nt er f ace0. . n 0. . n0. . n 0. . nCont ainsLogicalI nt er f aces

ResourceElem ent

LayerNet wor k

SubNet wor k

ResourcePor t

ResourceCollect ion

Com poundResour ce

Net workAt om icNet workCom posit e

Net wor knet wor kAdm inist r at iveSt at e : I nt egernet wor kAlar m St at us : I nt egerisNet workO per at ional : Booleanlayer Rat es : SequenceO f St r ing

LAN

. . .

. . .

. . .

. . .

HasCom poundResources

. . .

. . .

. . .

. . .

HasNet wor ks

ResourceElem ent

Net wor kAt om ic

{O nly one of LogicalDeviceG overnsTPsand LogicalDeviceG over nsPipes}

Net wor kCom posit e

Net work

ResourceCollect ion

Com poundResource

PhysicalPor t

Resour cePor tisEdgeResourcePor t : Boolean

DeviceI nt er f ace

Pipe

Ter m inat ionPoint

LogicalDevice

0. . 1

1. . n

0. . 1

1. . n

HasNet wor ks

. . .

. . .

. . .

. . .

HasCom poundResources

1. . n

1

1. . n

1

PhysicalPor t sI nResourcePor t

. . .

. . .

. . .

. . .

PPor t BindsToDeviceI nt er f aces

. . .

1

. . .

1

T Ps I nR eso urc eP or t

. . .

. . .

. . .

. . .

TPBindsToDeviceI nt er f aces

1

. . .

1

. . .

TPsI nPipe

. . .

. . .

. . .

. . .

LogicalDeviceG overnsDeviceI nt er f aces

. . .

. . .

. . .

. . .

LogicalDeviceG overnsPipes

. . .

. . .

. . .

. . .

LogicalDeviceG over nsTPs

Resour ceElem ent

Net wor kAt om ic

{ O nly one of LogicalDeviceG over nsTPsand LogicalDeviceG over nsPipes}

Net workCom posit e

Net work

Resour ceCollect ion

Com poundResource

PhysicalPor t

Resour cePor tisEd geR es our ceP or t : Bo olea n

DeviceI nt er f ace

Pipe

Ter m inat ionPoint

LogicalDevice

0. . 1

1.. n

0. . 1

1.. n

HasNet wor ks

. . .

. . .

. . .

. . .

HasCom poundResour ces

1. . n

1

1. . n

1

PhysicalPor t sI nResour cePor t

. . .

. . .

. . .

. . .


. . .

1

. . .

1

TPsI nResour cePor t

. ..

. . .

. ..

. . .

TPBindsToDeviceI nt er f aces

1

. . .

1

. . .

TPsI nPipe

. . .

. . .

. . .

. . .

Log icalD evic eG ov ern sD evic eIn ter fa ces. . .

. . .

. . .

. . .

LogicalDeviceG overnsPipes

. . .

. . .

. . .

. . .

LogicalDeviceG overnsTPs

Resour ceElem ent

Layer Net workSubNet wor k

connect ionRat es : SequenceO f St r ingsubNet workType : I nt eger

Resour cePor t

Net wor kAt om ic

L AN

ResourceCollect ion

Com poundResource

Net wor kCom posit e

Net wor knet workAdm inist rat iveSt at e : I nt egernet workAlar m St at us : I nt egerisNet wor kO perat ional : BooleanlayerRat es : SequenceO f St r ing

. ..

. . .

. ..

. . .

HasCom poundResour ces

. . .

. . .

. . .

. . .

HasNet works

Phy sica lRes ou rce LogicalResour cePhysicalRe sou r ceSp ecLogicalResourceSpecCom po und Re sou rce Sp ec

ResourceSpecif icat ion

Com poundResour ce

Resour ceCharact er ist ic

Resour ceusageSt at e : I nt eger

Resour ceCharact er ist icValue

LogicalRes our ceCha ract er ist icCharact er ist ic

. . . . . .. . . . . .

PResour ceSuppor t sLResource

. . .. . . . . .. . .

LResSpecBindsToPResSpec

. . . . . .. . .

I nvolvedResour ceSpecs

. . .

. . .

. . .

. . .

. . .

HasResour ceSpecif icat ions. . . . . .. . . . . .

Specif iesResource

. . .

. ..

. . .

. ..

Com poundResour ceAspect s

1

. . .

1

. . .

Resour ceCharact er izedBy

1. . . 1. . .

ResourceTakesO n

1

. . .

1

. . .

Resour ceDescr ibedBy

Com poundResour ceSpec

ResourceFacingService

Resour ceSpecif icat ion

Cust om er FacingServiceResourceFacingSer viceSpec

LogicalResour ceSpecPhysicalResourceSpec

Product Specif icat ion

. . .

. . .

. . .

I nvolvedResour ceSpecs

. . .

. ..

. . .

. ..

. . .

HasResourceSpecif icat ions

. ... . . . ... . .

CFSer viceRequiresR FServices

1 . . .1 . . .

Specif iesResour ceFacingSer vice

1

. . .

1

. . .

RFSer viceSpecHasResour ceSpecs

. . . . . .. . . . . .


. . . . . .. . .

Pr odSpecRef erences

. . .

. . .

. . .

. . .

. . .

Pr oduct SpecDef inePRSpecs

Resour ceSpecVersionDet ailsr esour ceSpecVer ValidFor : Tim ePer iodpref er redVer sionToUse : St r ingm inVer sionFor Use : St r ing

Logic alResour c eSpecVersionSpecVer sion

Ent it ySpecif icat ion

PhysicalResourceSpec LogicalResourceSpec

Resour ceSpecVersionresSpecRevisionNam e : St r ingresSpecRevisionFor m at : St r ingresSpecRevisionNum ber : St r ingresSpecRevisionReason : St r ingresSpecRevisionSem ant ics : St r ingresSpecRevisionTim est am p : Tim eSt am presSpecRevisionValidit yPer iod : Tim ePer iod

Com poundResour ceSpec

Resour ceSpecif icat ion

. . . . . .. . . . . .


0. . n1

0. . n1

M odif iesResour ceSpec

. . .

. . .

. . .

. . .

HasResourceSpecif icat ions

M an agedTransm issionEnt it y

Tr ailTerm inat ionPoint

TrailConnect ion

Connect ionTer m inat ionPoint

PipeisUniDirect ional : Booleanoper at ingLayer Rat e : St r ing

Ter m inat ionPointdirect ion : I nt egervendor TPNam e : St r ing

DeviceI nt er f ace

Resour cePor t

PhysicalPor t

. . .. . . . . .. . .

Connect ionsI nTr ail . . .. . . . . .CTPsI nCTP

. . .

1 . . .1 . . .

TPsI nPipe

. . .

. . .

. . .

. . .TPBindsToDeviceI nt er f aces

1

. . .

1

. . .

TPsI nResourcePor t

. . .. . . . . .. . .


1

. . .

1

. . .

PhysicalPor t sI nResourcePor t

TrailTer m inat ionPointConnect ionTer m inat ionPoint

Term inat ionPointdirect ion : I nt egervendor TPNam e : St r ing

DeviceI nt er f ace

ResourcePor t

PhysicalPor t

. . .. . . . . .CTPs InCTP

. . .

1

. . .

1

. . .

TPsI nResour cePor t

. . .. . . . . .. . .PPor t BindsToDeviceI nt er f aces

1

. . .

1

. . .

PhysicalPor t sI nResour cePor t

Net wor kAt om ic

LAN

Resour ceCollect ion

Net wor k

Net workCom posit e

. . .

. ..

. . .

. ..

Has Net wo rks

LayerNet wor k SubNet work

1 . . .1 . . .

SubNet worksI nLayer edNet work

. . .

. . .

. . .

HasSubNet wor ks. . .


What is Important About Autonomics

� #�� ,��. � �� 9 ��,�� 9�� &��9(��9��(�� 0��

� +�� &�� 9 ��G

% H�� &�

% ��&��6�

% �1��$��

% /��'��1��

% �� )��


��(�� $�� '��&��

% ��(�� $I��1��

% �� $�� &�� 1��

J ��

J ��

Autonomics and PeopleA Partnership


Automated Productivity: Self-Configuration

��

• Provide effective call-center support for 138 offices• Establish effective and efficient software distribution mechanism

��

��

��

“We rely heavily on information technology to drive business expansion and new levels of productivity and efficiencies. The self-configuring systems management delivered by Tivoli software is not only indispensable, it has so far been a key enabling force that has helped us bring such business and IT imperatives to success.”

—Peng Jin Song, MIS Director, China Netcom Corporation Ltd

• IBM Tivoli Configuration Manager• IBM Tivoli Remote Control

• Support 4,000+ staff across 138 offices with only 38 IT staff• Self configuring function re-initiates software distributions that are not

complete and removes undesirable applications• Maintained IT staff-level even after an increase of 200 applications • Efficient control of entire IT infrastructure from one location


What is Autonomic Networking?


Autonomic Networking Definition

� �� $�� $��

� ��( ��1��$��

% �� &�� )��

% ��$0� ��&��(��1��&��

% �� &�� 1��&��

% :6�� &��1��)��

� ��(�� 1$��& ��& (��


Autonomic Networking Scenario

Cellular 2.5G

Service Delivery Platform

Application &Content Servers

Cellular 3G Mobile Broadband Wireless Access

Cable/DSL/Fiber

Core IP Network

Communication Gateway

Communication Gateway Communication

Gateway

Softswitch

Wireless Services Manager

��

��


Autonomic Behavior Has Multiple Contexts


Autonomic Computing Element

Managed Resource

Model-Based Translation Layer

Autonomic Manager

Observe

rEasonLearn

Policy and Context Server

Compare

Foundation(Finite State Machines)

Act

How to Make Something Autonomic

��

��


MBTL Using SNMP and CLI

Vendor Converters

Sensors for GatheringVendor-Specific Data

Effectors for Issuing Vendor-Specific

Commands

Semantic Model Converter

NormalizedXML Data

NormalizedXML Commands

Autonomic System

DEN-ng Information

Model

DEN-ng Derived

DataModels

Information Bus

1.1.1.1.1.1.1=34000

1.2.1.5.1.1.10=true

1.3.3.1.1.2=230000

SNMP Parser

interface FastEthernet0/1shutdown

CLI Command Factory

Semantic Objects in XML


So What Are We Doing About It?Here’s Some Work in My Lab


FOCALE Architectural Roadmap

Current State =Desired State?

Managed ResourceManaged Resource

Analyze Data and Events


Determine Actual State

YESYES

NONO

Define New Device Configuration(s)

Model-BasedTranslation


Autonomic Manager

ControlControl ControlControl

ControlControl

ControlControl

Policy Manager

��

Policy Manager

��

Context ManagerContext Manager

Ontological Comparison

Reasoning and Learning

ControlControl







YESYES

NONO




Autonomic Manager


ControlControl

ControlControl

Policy Manager

��

Policy Manager

��




ControlControl



Genesis Autonomics Architecture

� *$�� H�� &��F�

� � ��#�� :�&��

� ��(��)��/��(

� :7�F/:�� ((��)��


Genesis Architecture

KnowledgeRepository

Autonomic Management Platform

Dashboard

Autonomic Management Platform Console

HTTP

KnowledgeManager

AutonomicAgents

Sensor(s)

Effector(s)

AutonomicApplications

(e.g. RF Optimization)

Core Services(e.g., FM, PM, CM)

ConsoleSession

Management

PolicyRepository

PolicyServer

L&R AlgorithmRepository

Learning &Reasoning

Engine

OntologyRepository

SemanticReasoning

Engine


Implementation of Genesis

CDL AgentOMCR 1

TCP/IP

KPI / KQICalculation

Vendor converters

ThresholdMonitor

DrilldownAnalysis

JMSQueue

JMSQueue

JMSQueue

PolicyManager

Orinoco DB

Dashboard

PlatformAJAX

(TIBCO GI 3.3 ®)

GUI Backend

XML/HTTP

EJBORDBMS(Intersystems Caché ®)

EJB

Cachénative

(JBoss 4.0.4GA)

J2EE Application Server

JBoss Rules

JiBX

Linux(Gentoo distribution Kernel 2.6.x)

HTML, JavaScript, XML

J2SE JDK 6, XML

EJB EJB

EJB

CDL AgentOMCR 2

CDL AgentOMCR n…

JMS

Agent Framework

Entities

Orinoco Console


Genesis Model

GenesisModel File

Tooling: Strategy and Importance

Rational RoseModel File

Source Code

Genesis Compiler

Genesis Syntax Checker

Form Editor

Source Code Editor

CodeHighlighter

Rational RoseConverter

Genesis Model Builder





JavaCode Generator

Java SyntaxChecker

ModelNavigator

Model-Model Object Map

Genesis Compiler

Model-Model Object Map

Import Rose andTransform to Genesis

View/Edit Genesis Model

Export Model andTransform to Source Code

Open/SaveGenesis Model


Tools to Build Legacy Models


Legacy � Autonomics

Overall Innovation: To provide increased flexibility without having to retool legacy devices and systems


1. Metamodel-Driven Reverse Engineering Legacy Artifacts

� � ��% ��

��

� � ��% ��

�� !�!"��"�� "�� !#

� � ��$�% %� �� & ��

��

Extract

Extractor


2. Design Self-* Models

An Example of a Self-Healing Model

ValueFailureplausability : Integer

Tim ingFailure

isLate : BooleanProvisionFailure

Report Rejuvenate Recover

CommissionOm ission

Risk

degree : Integer

Detector

Analyzer

1..n

1

1..n

1

Computes

MediationStrategy

Different detector types modeled as subclasses

SoftwareAtomic

Mitigator

1..n

1..n

1..n

1..n

StrategyAppliedTo

Failure

1..n 1..n1..n 1..n

FailureDetectedBy

1..n 1..n1..n 1..n

FailureAnalyzedBy

1..n1..n 1..n1..n FailureSemanticsSuppliedTo

Software

1..n 1..n1..n 1..n

AppliesMitigationTo

0..n

0..n

0..n

0..n

HasFailure

SoftwareComposite

0..n

0..1

0..n

0..1

HasSoftware


3. Base Model and Self-* Model Integration Using Aspect-Oriented Modeling (AOM)

� %��'(��

� )��'(�� * �� * � �� '(��

� )+ & �� % )�� ,� ��

��"��)�� * �� #�� (�� #

% ! �)��

� -�� .'�)/ "�& � ��/ %)0�"�)& / "�1�� "�2


4. Base Code and Self-* Code Integration Using Aspect-Oriented Programming (AOP)


Managing Using State Automata


Traditional Finite State Machine Theory

� 3�� )� ��

� & ��& ��& ��

� ��4� � � ��

� ��5

� 6 & 5�� & ��

� 2

None of these relates a state to the information that it represents!

Traditional FSMs don’t consider semantics of behavior!


State Machine Models

Policy-driven Behaviour Orchestration Using Motonomics State Machine

S

A

B

C

D

T

2

44

33

f1(x)

Policy1

f2(x)

Policy2

Structural Models

attr1=2, attr2=3

verification ofactual state

Policy3

f3(x)

attr1=3, attr2=4policy determinesdesired state

attr1 : int

Foo

attr2 : int

optimal path


Adaptive Control Functions






YESYES

NONO




Autonomic Manager


ControlControl

ControlControl

Policy Manager

��

Policy Manager

��




ControlControl







YESYES

NONO




Autonomic Manager


ControlControl

ControlControl

Policy Manager

��

Policy Manager

��




ControlControl



Policy Management


Business View: SLAs, Processes, Guidelines, and GoalsBusiness View: SLAs, Processes, Guidelines, and Goals

System View: Device- and Technology-Independent OperationSystem View: Device- and Technology-Independent Operation

Administrator View: Device- Independent, Technology-Specific OperationAdministrator View: Device- Independent, Technology-Specific Operation

Device View: Device- and Technology-Specific OperationDevice View: Device- and Technology-Specific Operation

Instance View: Device-Specific MIBs, PIBs, CLI, etc. ImplementationInstance View: Device-Specific MIBs, PIBs, CLI, etc. Implementation

The Policy Continuum


*� ��1 ��(��/"�� "F�� ((��&�1��

� ��

��

"��

*� ��(�� $(�� 8��&��1 ��(��)�� (�1��

��!��

��

� ��"��

*�)��

*� �� ((��&�1�� )��*� ��)9� ��*� ��)9� ��

#��$��%��

��& ��%��'�(��)%* ��

�� "��

�� )�

�� 9 ��&��)��KL ��)��F��0�M

$��%��$��

�� N�L ��)��N�L ��((��

+��& ��%��F��

�� (��5 1 �� (��5 1 ��)��

Mapping Examples


How Does Policy Fit In?

ContextRelevant

Policy

Autonomic ManagerContext Manager

��


Context-Aware Policy Management






YESYES

NONO




Autonomic Manager


ControlControl

ControlControl

Policy Manager

��

Policy Manager

��




ControlControl







YESYES

NONO




Autonomic Manager


ControlControl

ControlControl

Policy Manager

��

Policy Manager

��




ControlControl



Knowledge Management


Diverse Knowledge Representations �� '��77

��8��# �� 8��98#

��8��8�8��##:; �� 8��###

�� '�� 77 �� $<##8��=' ��8��# ��8��8��# ��* ��8�8 �� 8��#:; � �� 8�# �� 7> �> �> �7�8 �� 8��##PolicyMetaData

priority : Integer = 0

PolicyRuleComponentStructure

PolicyRuleComponentMetaDatapolicyComponentRuleConstraints : StringpolicyComponentValidFor : TimePeriod

0..1

1..n

0..1

1..n

PolicyRuleComponentHasMetaData

PolicyRuleStructure

PolicyRuleMetaDataisMandatoryEvaluation : Boolean = TRUEpolicyRuleActionConstraints : StringpolicyRuleEvalConstraints : StringpolicyRuleValidFor : TimePeriod

0..n

1..n

0..n

1..n

PolicyRuleHasMetaData

PolicyEventPolicyCondition

PolicyEventStructurePolicyConditionStructure

ECAPolicyRulehasSubRules : Boolean = FALSEisCNF : Boolean = TRUE

0..n1..n

0..n1..n

ECAPolicyRuleUsesPolicyEvent

0..n1..n

0..n1..n

ECAPolicyRuleUsesPolicyCondition

PolicyActionStructure

0..n1..n

0..n1..n

ECAPolicyRuleUsesPolicyAction

PolicyAction


Model-based Knowledge Engineering & Transformation

GenesisGenesis

DEN-ng models: Info, data, policy, behavior

DEN-ng models: Info, data, policy, behavior Ontology ModelOntology Model

Knowledge modelKnowledge model

KBsKBs

OntologiesOntologies

Generated codeGenerated code

Model transformationModel transformationMLML


Purpose of Machine Learning

� #��1��$�� (��)��1��6(��

� #��&�� 6��&�(��(�� &��&�9��)��$�

� #�� $�� I�'�� &��1�


Reinforcement Learning� ��&�&��

��)�� % 5 1��)��)�� I��,(��.��

% #'��

% ��)��,�� .��

� �&�� &9��&��)�� 0 �"�#

�� & �� )�� O

% ��/��&�� 9��)��

� �/��2��9��9��3 ��&�� )��

� 7��$��&9��&��$�� 8�� ,��&�� &��$.

��

��

!��"��

��

�∞

= ++=0 1k ktr

ktR γ

{ }( )),()','(),( asVasVrasV −+=∆ γα


Purpose of Reasoning

� �� $(��$(�� $�� &��

� � �� '�� $(��$��

% F��,�� $(��$�(��)��)��$

� ��&�� $�(��)��,�,�� - ��'�� &��1�


Semantic Equivalence

Device M…Command 82Command 83…Command 93Command 94Command 95…Command j

1.0

Device A…Command 64Command 65Command 66Command 67…Command i

BGP Peering

0.95

OntologyDifferences Effects from Device A

Effects from Device B

To Policy Server

� "� �� *:7 9�&"� �4��

� ��(��5 /� ��&��&


Machine Learning Model Selection

? ��3��@��? ��? ��)��(��

@��? ��3��@��@��"��(��1��$

3��@��@��? ��? ��6��

@��? ��? ��@��@��1��$��)��(��

@��? ��? ��3��@�� (��1��$

? ��@��? ��@��@��1��

? ��? ��? ��@��@�� &�)��

@��? ��? ��@��@��7��&�� 2� �6��3 �$(�

H77� ��#��77��

There exist various learning models, each of which has strength and weakness.The model selection always relies on some guidelines, expertise and empirical studies.

In an autonomic computing framework, the selection needs to be automated!


RL based learning selectionProblem: Impact Prediction for a BTS controller Failure

Solution: Neural Network predictive models that are proven to beuniversal function approximators

Decision Making Criteria:• A relatively stable environment prefers a global approximator• A relatively dynamic environment prefers a local � ��

Multi-Layer Perceptron (global) Radial Basis Function (local)


Global vs. Local – No Universal Answer� Offline use provides very similar performance with proper

training and validation� Online use reveals significant differences under time constraints

• MLP is less sensitive to change in its learning data• RBF is much more sensitive to change because of its use of

unsupervised learning in the input space

MLP (global) RBF (local)


Summary


It is NOT “Just” About Technology!


Networking Scenario – Service Maintenance

�#��$ �%&�'


Is Harder Than It Looks…

()*+,,-.

�#�� )��/'


It’s a True Paradigm Shift

� 2:)��$��&��1�� (��(��1�� .

� ! �� 1�� $�� (�� '��$�� 1��1��)��&��

��

�� !��


Questions?Questions?

“Create like a god. Command like a king. Work like a slave”- Constantin Brancusi


Autonomic Networking References

� ��(O44��&4�� 9�� '��&4

� ��(O44� � � ��9(�� &4

� ��(O44� � � ��9(�� &4

� ��(O44� � � ��&&��(�� &4��6�(�(4� ��P�&�

� ��(O44� � � �1�$��9��9��0��4��(6G"�Q?D"EE"?A=">?>"BBE>>E

� ��(O44� � � ��1��4�� AC4

� ��(O44� � � ��1��''� �4"! ��?AAC4

� ��(O44(�� 4��

� ��(O44��(��1�� 6��4R��AC4

� ��(O446�&�� 9��&��'4�5 *�?AAC4

� ��(O44� � � �� 4R��?AAC4


Conferences and Web Sites

� "�� (��&�,"��?AAC.% ��BB9B@��?AAC��'��)��% ��(O44� � � �� 4R��?AAC4% �� (��&�� (�� "��IA=�� IA<% ��)�� '��(�� "��?AAC

J +��#�(�� (��&J ��(��)�� $�� J :�&��&�:� ��&��*��0�� $�� J S �� 7�� 'J ��$9F�� (��&

� ! �1��% L ��O�� (��&��&% L ��O�� 1� �� 4�� % ��O��


Autonomic Computing References

� TBU�� (��&��((��

"F� I�� O�� 1� �� 4�� 4� ��

H�(��5 ��*�� 2#�� (��&3��":::�� (��$�?AA>�� 1� �� 4�� 4��4((��4��P��P�� (��P��P?AA>�(�

H�(��5 ��2��&�� (��&3��&�� ?C��"�� :�&��&�,?AA@.% �� )��@A�� (��&��((��

� T?U�"F� I�5 ��*�� (��&��)�

��(O44� � � C1�� 1� �� 4�� 4��1��$4��4A>A?�� 14A>A?�� 1��

��(O44� � � 9BA<��1� �� 4��)��(�� '41�� 4��=��

��(O44� � � �� 4� ��4� ��((��4�1��(� G�� :��"5 7Q75

� T>U�� $�� I 7B��)��(O44� � � �� 4� �� 4��4�B4

� T=U�+ �I ��(��)��"� ��)��(O44� � � ��(�� 4��&�4&��1��4��


Foundational References (1)TBU ��2��7�� &� � ��7�� '�� &�� F��*��)��*�)��

� �&�� 3��%%& )) /00/��

T?U ��$��1�� 2�� !� � �� #��1��&��BDDD��"�F7 �B9@CECA9B=A9<

T>U ��2��$9F��7�� '�� &�� 3�� &��H� � ��1��(�?AA>��"�F7 �B9@@E<A9E@D9B

T=U ��'��+ ��+ ��#�� 1��&��)��$��BDD<T@U ��#� ��H��2� '��&�7L 5 �� )��&�� (��6��$3��

#:�+ ?��#� ! ��7�� $�?AA=�T<U ��2F��&�F��#�� #��&�� /� ��

�$��3��:�� 7 �?AA=��5 ��?AA=�TCU ��F�� 2�� $�"� �� "�� $��

�� (��&��$�� 3�� *�"�?AA<TEU ��2�� 0��&�� 9L �)�� $��

��7�� '�2��:�� ?AA<�H�$��TDU ��2� ��&�� )�� &��)��&��)��7�6��L ��

7�� '3��'�$��&�)��"��*�1��A<TBAU ��7��&�� :��/��2�5 ��/:�� 7�)�� (��&�

��3��/��?AA<�, ��/�� (��&�� .��$�?AA<�,�6��)��"#�� ?AAC.

TBBU ��7��&�� :��/��25 ��&$9F��H�� &��(�� 9L �)��&��$�� 3��*�5 � �?AA<

TB?U ��2��$9F��6�9�� )�� &�� '� �� (��&3��"7#://�5 � �?AA<

TB>U 7��&�� F��1�� *��F��)��:��/��! �*��$��2��&�� 7�� '�� &�� 3�� :�?AA<


Foundational References (2)

� ��(O44� � � 9'�� 4'�4� ��9�9�9��&$��

� ��(O44� � � ��9��4&��

� � ��#�� /��&��7�� V��'��7VO�� L �� 9+ ��"��:��BDDC��((��?A9=@�

� ��(��W ��(��1��)��"� ��O�� (��(�$��#��&$�� 1��&�� HO�� 1��&��)��$��BDD<��C�


MDA, Patterns, and RolesTBU�F�� $��25 1 ��95 ��

��3��9+ ��"�F7 �A9B>9<?DB@@9=��BDDC

T?U�� 6� (�� 2��$��1��5 1 �� 3��"�F7 �A9?AB9ED@=?9A

T>U�"��(��)�� 1 ��(�� (��(O44� � � ��&4�� (��9��9��4BDDC4(��(9BDDC9��9�1 ��(�


Framework References

TBU��O�� &��&4� �

T?U5 � L ��2�� &�/�&�&��(�� 3��)��B�@�� ?AA>

T>UX�� 'O��O�� 0� ��

T=U5 (��*��1��&�� "�5 4":��BAC=<9?��BDD<

T@U��*��#�� *�)��(� ��"��1��L ��F�� 1�&��9! ��$�BDDD


Applications

� *F?�� #��&�,*F?�)E�B.

% ��(O44� � � ��1� �� 4 ��4�1?4��(�(4>>>DA=B

% Y#��&�� "F� �*F?�� )��*�1��)��Y �� $��'��H�)��V�6�� *��&��(��/��+ �� /�&�� + �� ?AA=��

� �� 9��&�� $�� &��

% Y� ��&�� "� 5 �/�� /��F��& �� (��&��$�� Y V�6�� *��(��/��+ �� /�&�� $ ��'��L ��9�� ?AA=��

% 2"��(��&�� "��*��&�� /��F��&��3 V�6�� *��(��/��+ �� /�&�� $ ��'��L ��9��9#�� :� 1��#��&$��((��$�� $� (�� ?AA=��

Documents

Autonomic Administration - HAL 9000 Meets Gene … · Autonomic Administration: HAL 9000 Meets Gene Roddenberry ... DCN DRS CPNI Sales Comp PB Awards ... Paging CNR PBITS MP/F FLEXCOM