autonomic computing meets complex information · pdf fileautonomic computing meets complex...

Prof. A. TalebProf. A. Taleb--Bendiab, Keynote Lecture: WEBISTBendiab, Keynote Lecture: WEBIST’’07, Barcelona, Date: 13/09/2007, Slide: 07, Barcelona, Date: 13/09/2007, Slide: 11

Autonomic Computing Meets Complex Information Systems:

Theory and Practice

Prof. A. Taleb-BendiabSchool of Computing

Liverpool John Moores Universityemail: [email protected]://www.cms.livjm.ac.uk/taleb

Autonomic Computing Meets Autonomic Computing Meets Complex Information Systems: Complex Information Systems:

Theory and PracticeTheory and Practice

Prof. A. TalebProf. A. Taleb--BendiabBendiabSchool of ComputingSchool of Computing

Liverpool John Moores UniversityLiverpool John Moores Universityemail: email: [email protected]@livjm.ac.ukhttp://http://www.cms.livjm.ac.uk/talebwww.cms.livjm.ac.uk/taleb


OutlineOutlineUnderstanding the gap between theory and practice Understanding the gap between theory and practice

Pervasive and Trusted Network and Service Infrastructures Drivers for a paradigm shift Drivers for a paradigm shift ---- Software Intensive SystemsSoftware Intensive Systems

In the quest of better In the quest of better ““Service and software architectures, infrastructures and engineering” modelsmodels

Recent bRecent bioio--inspiredinspired initiativesinitiativesSAS, DASADA, SRS and ANTSSAS, DASADA, SRS and ANTS

A brief Introduction into Autonomic A brief Introduction into Autonomic systems engineeringsystems engineeringDefinitions and state of the art Definitions and state of the art What has been done so farWhat has been done so far

Challenges and Open Questions Challenges and Open Questions

Conclusions and Q&AConclusions and Q&A


Emerging Networked LandscapesEmerging Networked Landscapes


Drivers for a Paradigm Shift #1Drivers for a Paradigm Shift #1Modern ExpectationsModern Expectations

High-Availability -- 24x7 delivery nearnear--100% availability is becoming mandatory for 100% availability is becoming mandatory for ee--commerce, enterprise apps, online services, ISPscommerce, enterprise apps, online services, ISPs

ChangeSupport rapid deployment of new hw/sw, services, etc

MaintainabilityProvide flexible systems admin. env.

reduce system administrators tasks, complexity and costJust-in-time scalability

Allow flexible system up scaling without sacrificing performance, availability or maintainability

evolutionary growth and adaptationSurvivabilityFull malleability


Drivers for a Paradigm Shift #2Drivers for a Paradigm Shift #2Key question is not only how to achieve the above listed modern expectations as:

a single metric/attribute or a cost-effective combination of them all

Within multi-service provider settings and management domains

how to reduce the cost and complexity of achieving that

HowNature-inspired models – Autonomic Computing

Management by delegationEtc.


Some Recent InitiativesSome Recent InitiativesSAS SAS ---- SelfSelf--Adaptive Systems Adaptive Systems (DARPA, 1997)(DARPA, 1997)DASADADASADA -- Dynamic Assembly for Systems’ Adaptability, Dependability, and Assurance (DARPA, 2000)DARPA, 2000)AC (IBM, 2001) AC (IBM, 2001) Autonomic Communication (EU, Autonomic Communication (EU, 2003)2003)SRS SRS ---- SelfSelf--Regenerative Systems Regenerative Systems (DARPA, 2003)(DARPA, 2003)ANTS ANTS ---- Autonomous Autonomous NanoNano--Technology Swarm (NASA) Technology Swarm (NASA)

Major Recent InitiativesMajor Recent Initiatives

Defined by Defined by LaddagaLaddaga in the in the 1997 DARPA Broad Agency 1997 DARPA Broad Agency Announcement as:Announcement as:

““...software that evaluates ...software that evaluates its own performance and its own performance and changes changes behaviourbehaviour when when

the evaluation indicates that the evaluation indicates that it is not accomplishing what it is not accomplishing what the software is intended to the software is intended to

do...do...””..

To adapt, the system reacts To adapt, the system reacts to environmental change to environmental change -- the the

problem is problem is recognisingrecognising the the need for change, then need for change, then planning, enacting and planning, enacting and

verifying the change verifying the change -- these these are selfare self--managing concernsmanaging concerns

•• More on the IBM More on the IBM Autonomic Computing Autonomic Computing later onlater on••For Autonomic For Autonomic communication See M. communication See M. Smirnov Smirnov pptppt..

4 major research thrusts: 4 major research thrusts:

•• BiologicallyBiologically--inspired inspired diversitydiversity: : genetically diverse genetically diverse

computing fabriccomputing fabric•• ““Cognitive immunityCognitive immunity”” and and selfself--healing: see automated healing: see automated cyber immune response and cyber immune response and

system regeneration. system regeneration. •• Granular, scalable Granular, scalable

redundancy: This research redundancy: This research thrust area will increase the thrust area will increase the practicality of redundancy practicality of redundancy

techniques.techniques.•• Reasoning about the

insider threat to preemptinsider attacks and detect

system overrun.

A 2020 vision of a class of A 2020 vision of a class of space exploration missions space exploration missions termed termed nanoswarmsnanoswarms, , where where many cooperating many cooperating picospacecraftpicospacecraft or intelligent intelligent spacecraft work in teams to spacecraft work in teams to explore the asteroid belt, explore the asteroid belt, based on the efficiency and based on the efficiency and coordination of hive culture.coordination of hive culture.


Theoretical BackgroundTheoretical BackgroundHas been informed by a set of design paradigms Has been informed by a set of design paradigms

ModelModel--based based vsvs SelfSelf--Organising Systems design modelsOrganising Systems design modelsTopTop--down down vsvs bottombottom--upup

Applying and/or revisiting:Applying and/or revisiting:cybernetic principlescybernetic principles

control systems theory, regulation, reward and control systems theory, regulation, reward and sanctionssanctions

Decision theory, Complexity theoryDecision theory, Complexity theoryDAI and CIDAI and CI

dynamic planning, deliberative models, MLdynamic planning, deliberative models, MLMiddleware supportMiddleware support

selfself--awareness, reflection and deliberationawareness, reflection and deliberationAutonomic Software ArchitectureAutonomic Software Architecture, etc., etc.


Definitions #1Definitions #1CCan be defined as:an be defined as:

3. Yet Another3. Yet Anothercan operate independently of (or with can operate independently of (or with limited) human intervention, thus hiding limited) human intervention, thus hiding their systemstheir systems’’ design and management design and management complexity including intricacies of the complexity including intricacies of the automation of laborious administration tasks, automation of laborious administration tasks, recovery from unanticipated systemrecovery from unanticipated system’’s failure, s failure, and/or selfand/or self--protection from security protection from security vulnerability. vulnerability.


CharacterisingCharacterising AC CapabilitiesAC CapabilitiesCharacterisingCharacterising AC SystemsAC Systems

A software system is A software system is autonomic, if it possesses the autonomic, if it possesses the following capabilities:following capabilities:

SelfSelf--configuringconfiguring —— choosing a choosing a suitable suitable behaviourbehaviour, based on , based on user preferences, context, user preferences, context, ……SelfSelf--tuningtuning —— choosing choosing behavioursbehaviours that optimize certain that optimize certain qualities (performance, yearqualities (performance, year--end end profits, profits, ……))SelfSelf--repairingrepairing —— shifting shifting execution to another execution to another behaviourbehaviour, , given that the current one is given that the current one is failingfailingSelfSelf--protectingprotecting —— choosing a choosing a behaviourbehaviour that minimizes risks that minimizes risks (attacks, viruses, (attacks, viruses, ……))


The The Story So Far Story So Far #1#1StateState--ofof--thethe--art of autonomic systems designs including;art of autonomic systems designs including;

autonomic software models and architecture, autonomic software models and architecture, standards standards tools and techniques to support tools and techniques to support

the design, modelling, analysis the design, modelling, analysis and evolution of autonomic software and evolution of autonomic software

Define associated models for their Define associated models for their programming, control programming, control interaction models with human and/or other noninteraction models with human and/or other non--AC systems (legacy). AC systems (legacy). Delegation of authority, Delegation of authority,

its adjustment and revocationits adjustment and revocation


Open Standards

The Story So Far The Story So Far #2#2

Autonomic core capabilities

Products delivering autonomic features

Business policy

Aut

onom

ic C

ompu

ting

Arc

hite

ctur

e

Solution Install

Problem Determination Admin Console

Policy

Resource Provisioning

• Define a base reference architecturemodel which creates a common vernacular for autonomic computing

• Deliver core infrastructure technologies that provide for an open framework for the industry• Deliver products with built-in autonomic capabilities

• Create and leverage open standards for autonomic computing

Workload Management


Remaining Research ChallengesRemaining Research ChallengesHostHost--BasedBased

Complexity ParadoxComplexity ParadoxAutonomic computing aims to reduce admin. costs, hide system comAutonomic computing aims to reduce admin. costs, hide system complexity and plexity and intricacy, intricacy, Though, their designs are becoming more complex Though, their designs are becoming more complex

as yet are poorly understood as yet are poorly understood as echoed by D. Garlan 2005as echoed by D. Garlan 2005“…“… how do we design, build, and evolve such sw systems so that thehow do we design, build, and evolve such sw systems so that they y can meet givencan meet given——and evolvingand evolving——requirements ...requirements ...””

Incremental deployment of AC capabilities in legacy Incremental deployment of AC capabilities in legacy systems. systems. AOPAOP--based evolution, Interoperationbased evolution, InteroperationSupport functional and nonSupport functional and non--functional requirements for functional requirements for autonomy.autonomy.Evaluation mechanisms and metrics [ref] Evaluation mechanisms and metrics [ref]

Governance Governance vsvs Autonomy ParadoxAutonomy ParadoxBalancing and adjusting governance and autonomyBalancing and adjusting governance and autonomyProgramming, control and Interaction Models Programming, control and Interaction Models

Usage paradoxUsage paradoxEnriching and Interacting with BPEnriching and Interacting with BP

ProcessProcess--centric interactioncentric interactionComplexComplex--BasedBased

SelfSelf--organisationorganisationComplex and random Networks of autonomic systemComplex and random Networks of autonomic system


The Story so far The Story so far #3#3More recent work is focusing on scalable More recent work is focusing on scalable methods for specifying methods for specifying

dynamic behaviour of autonomic systems. dynamic behaviour of autonomic systems. Evolving Evolving

policies and control modelpolicies and control modelStructural/organisational modelStructural/organisational model

Bounded autonomy and adjustmentsBounded autonomy and adjustmentsUnifying models for Unifying models for

modelmodel--based and SOS approaches for autonomic based and SOS approaches for autonomic systems engineering and managementsystems engineering and management


Designing Autonomic Systems #1Designing Autonomic Systems #1More recent theoretical work is focusing on scalable methods forMore recent theoretical work is focusing on scalable methods forspecifying and enacting dynamic behaviour of autonomic systemsspecifying and enacting dynamic behaviour of autonomic systems

Autonomic Systems EngineeringAutonomic Systems EngineeringRelated Work: Related Work: IBMIBM blueprint blueprint (www(www--03.ibm.com/autonomic/pdfs/ACBP2_200403.ibm.com/autonomic/pdfs/ACBP2_2004--1010--04.pdf)04.pdf)

An autonomic manager contains a continuous control loop that monAn autonomic manager contains a continuous control loop that monitors activities and takes actions to adjust the itors activities and takes actions to adjust the system to meet business objectivessystem to meet business objectivesAutonomic managers learn from past experience to build action plAutonomic managers learn from past experience to build action plansansElements need to be instrumented consistently, based on open staElements need to be instrumented consistently, based on open standardsndards

Our modelOur modelModelModel--based Approachbased Approach

Systems theory, design patterns, design grammar and serviceSystems theory, design patterns, design grammar and service--oriented programmingoriented programmingA. TalebA. Taleb--Bendiab, D.W. Bustard, R. Sterritt, A. Laws, M. Randles, F. KeenBendiab, D.W. Bustard, R. Sterritt, A. Laws, M. Randles, F. Keenan, P. Miseldine, "an, P. Miseldine, "ModelModel--Based Based SelfSelf--Managing Systems EngineeringManaging Systems Engineering", in Proceedings of the 16th International Workshop on Database", in Proceedings of the 16th International Workshop on Databaseand Expert Systems Applications (and Expert Systems Applications (DEXADEXA’’0505), ), SAACS'05SAACS'05: 3rd International Workshop on Self: 3rd International Workshop on Self--Adaptable Adaptable and Autonomic Computing Systems, pp., and Autonomic Computing Systems, pp., David Bustard, Roy Sterritt, A. TalebDavid Bustard, Roy Sterritt, A. Taleb--Bendiab, A. Laws, M. Randles, F. Keenan, 05, "Bendiab, A. Laws, M. Randles, F. Keenan, 05, "Towards a Systemic Towards a Systemic Approach to Autonomic Systems EngineeringApproach to Autonomic Systems Engineering", ", EASE'2005EASE'2005. . David Bustard, Roy Sterritt, A. TalebDavid Bustard, Roy Sterritt, A. Taleb--Bendiab, A. Laws, M. 06, "Autonomic System Design Based on the Bendiab, A. Laws, M. 06, "Autonomic System Design Based on the Integrated Use of SSM and VSM", to appear in AI Review, Vol. , NIntegrated Use of SSM and VSM", to appear in AI Review, Vol. , No. , Springer, ISSN 0269o. , Springer, ISSN 0269--2821.2821.K. LiuK. Liu, , A. TalebA. Taleb--BendiabBendiab, 05, ", 05, "Presenting a Case for a Principled Approach to Presenting a Case for a Principled Approach to Citizen, Citizen, Business and Technology IntegrationBusiness and Technology Integration in ein e--Government Services: Challenges and Government Services: Challenges and Research OpportunitiesResearch Opportunities", ", Egov'05Egov'05. .

Monitor Execute

Analyze Plan

Knowledge

Managed Element

Sensors Effectors


Design Patterns: VSM and GOFDesign Patterns: VSM and GOF


Outlined in the Autonomic computing Outlined in the Autonomic computing edge: The edge: The ““StandardStandard”” way of autonomic way of autonomic computing, Brent Miller, 29/05/2005.computing, Brent Miller, 29/05/2005.Though more is requiredThough more is required

Standards Standards ---- InteroperationInteroperation


AC Design Process AC Design Process ---- correct correct ababinitioinitio

Classical SE approaches Classical SE approaches TopTop--down down ---- right 1st time right 1st time ----

Ecosystem ApproachEcosystem ApproachComplexity: Complex SystemsComplexity: Complex Systems

Seth Bullock and D. Cliff (HP Report, Ref.)Seth Bullock and D. Cliff (HP Report, Ref.)

Complexity and Emerging Behaviour in IT Complexity and Emerging Behaviour in IT SystemsSystems


Entropic Autonomic DesignEntropic Autonomic Design

…Printer Internet Other Facility or Service

…Application 1

Application 2

Application n

A2: monitor that the computing service facilitates business activities and take control action as necessary

A8: install new

technology A10: withdraw

obsolete technology

A7: maintain record of

technology in use

A4: maintain stock of spare equipment, parts and supplies

A9: procure new technology

A5: be aware of the cost of the

computing service

A3: maintain technology

in use A1: provide a computing

service

A6: be aware of the benefit of the

computing service TCA

A2: monitor that the computing service facilitates business activities and take control action as necessary

A8: install new

technology A7: maintain

record of technology in use

A4: maintain stock of spare equipment, parts and supplies

A5: be aware of the cost of the

computing service

A15: develop technology support

plan

A10: withdraw obsolete

technology A3: maintain technology

in use

A6: be aware of the benefit of the

computing service

TCA

A0: Provide a computing service

A1: provide a user service

A11: be aware of co-ordination needs

and implement controls

A12: be aware of technology opportunities

& threats

A13: be aware of organisation

needs

A14: be aware of organisation constraints

A9: procure new technology

A16: define expectations


Modelling Autonomy #1Modelling Autonomy #1Algebraic Algebraic specification specification

Process AlgebraProcess AlgebraCSP CSP Static model checking and dynamic software analysisStatic model checking and dynamic software analysis

Key States and Transitions are not enoughKey States and Transitions are not enoughIntelligent organisation emerges as in nIntelligent organisation emerges as in natural systemsatural systems

Our model Our model ---- axiomaticaxiomaticUsing SSC used to formalizes the behaviour of dynamically changiUsing SSC used to formalizes the behaviour of dynamically changing ng systems FOL (McCarthy, 1963).systems FOL (McCarthy, 1963).

Support concurrent actions and timing constraints. Support concurrent actions and timing constraints. Each situation can be viewed as a history of previous actions. Each situation can be viewed as a history of previous actions. Action, guards and time can be modelled at deliberation points iAction, guards and time can be modelled at deliberation points in n an autonomic setting.an autonomic setting.

M. Randles, A. TalebM. Randles, A. Taleb--Bendiab, Philip Miseldine, Andy Laws, "Bendiab, Philip Miseldine, Andy Laws, "Adjustable Adjustable Deliberation of SelfDeliberation of Self--Managing SystemsManaging Systems", ", ECBS 2005ECBS 2005: 449: 449--456. [456. [pptppt]]


Modelling Autonomy #2Modelling Autonomy #2Evolving and Adjustable AutonomyEvolving and Adjustable Autonomy

Via compositional model and evolution of Via compositional model and evolution of Software services (components)Software services (components)Software governance (Control) viaControl) via

Formal modelling of norms, policiesFormal modelling of norms, policiesEnactment support Enactment support –– from spec. to code from spec. to code using using Neptune language Neptune language M. Randles, A. TalebM. Randles, A. Taleb--Bendiab, P. Miseldine, 05, "Bendiab, P. Miseldine, 05, "Mind out of Programmable Mind out of Programmable Matter: Exploring Unified Models of Emergent System Autonomy forMatter: Exploring Unified Models of Emergent System Autonomy forCollective SelfCollective Self--Regenerative SystemsRegenerative Systems", Extended Abstract, the 2nd ", Extended Abstract, the 2nd GSFC/IEEE Workshop on Radical Agent Concepts (GSFC/IEEE Workshop on Radical Agent Concepts (WRAC'05WRAC'05), NASA GSFC ), NASA GSFC Visitor's Center, Greenbelt, MD, 20thVisitor's Center, Greenbelt, MD, 20th--22nd September 2005. 22nd September 2005. Miseldine, P., TalebMiseldine, P., Taleb--Bendiab A. Bendiab A. ““A Programmatic Approach to Applying A Programmatic Approach to Applying Sympathetic and Parasympathetic Autonomic Systems to Software DeSympathetic and Parasympathetic Autonomic Systems to Software Designsign””, , SelfSelf--Organisation and Autonomic Informatics (ISBN IOrganisation and Autonomic Informatics (ISBN I--5860358603--577577--0), Hans 0), Hans CzapCzap, Rainer , Rainer UnlandUnland, Cherif , Cherif BrankiBranki, , HuagloryHuaglory TianfieldTianfield (Eds.), pp:3(Eds.), pp:3--17, IOS 17, IOS Press, Amsterdam, 2005.Press, Amsterdam, 2005.Awarded "Most Innovative Paper" at ConferenceAwarded "Most Innovative Paper" at ConferenceMiseldine, P., TalebMiseldine, P., Taleb--Bendiab, A., Bendiab, A., ““CACA--SPA: Balancing the Crosscutting SPA: Balancing the Crosscutting Concerns of Governance Autonomy in Trusted SoftwareConcerns of Governance Autonomy in Trusted Software””, , IEEE IEEE International Workshop on Trusted and Autonomic Computing SystemInternational Workshop on Trusted and Autonomic Computing Systems s within AINA 2006. Vienna, Austria. April 2006.within AINA 2006. Vienna, Austria. April 2006.


Governing AutonomyGoverning AutonomySelfSelf--Governance TheoryGovernance Theory

Formal modelling of norms, etiquette, rules of playFormal modelling of norms, etiquette, rules of playdeployed via CAdeployed via CA--SPA constructs SPA constructs Enactment support Enactment support –– from spec. to codefrom spec. to code


Learning Control Rules as an Learning Control Rules as an Emerging Behaviour Emerging Behaviour


Emergence of Control Rules Emergence of Control Rules ––Adjustable Autonomy in ActionAdjustable Autonomy in Action

To run the AVI of Soot in To run the AVI of Soot in action, click inside the action, click inside the window below.window below.

This shows how Neptune This shows how Neptune language is used to language is used to generate and deploy a generate and deploy a metameta--system observer, system observer, which can observe the selfwhich can observe the self--organising soot algorithm, organising soot algorithm, and learn autonomic control and learn autonomic control rules from the system rules from the system operation.operation.The control rule as action The control rule as action history are randomly shared history are randomly shared with other soot agents and with other soot agents and rewarded the more they are rewarded the more they are used used –– AKO distributed AKO distributed reinforcement learning.reinforcement learning.This approach supports This approach supports control rules extraction too. control rules extraction too. Etc.Etc.


BP to SOA BP to SOA ---- Dental Triage Demo.Dental Triage Demo.

Current System

Re-engineerin

g via

Neptune to

olkit

New

Visual Modelling of Protocol or process flow

New Grid-Based System

Process is compiled into an

open introspective format


2nd Scenario: Situation2nd Scenario: Situation--Aware Decision Aware Decision SystemsSystems


ProcessProcess--Oriented ProgrammingOriented Programming#1#1


ProcessProcess--Oriented Oriented Programming Programming #2#2


ProcessProcess--Oriented Programming Oriented Programming #3#3


PetshopPetshop BenchmarkBenchmark


Neptune and Neptune and DIY DIY ComputingComputing #1#1

ApplicationApplication--centric centric mashupmashupUser requirements User requirements ---- processprocess

Our approach Our approach ---- NeptuneNeptuneAllows for process descriptions to be mapped, Allows for process descriptions to be mapped, computationally, to actual source codecomputationally, to actual source code

Developers can write codeDevelopers can write codeDomain experts can write ontological and domainDomain experts can write ontological and domain--specific specific modelsmodels

Fully open, and exportableFully open, and exportableXML basedXML basedSuited for current social network toolsSuited for current social network tools


Neptune and Neptune and DIY DIY ComputingComputing #2#2

WikiWiki tools etc. can be used to produce tools etc. can be used to produce useruser--friendly, centric services to once friendly, centric services to once complex, and difficult computational taskscomplex, and difficult computational tasks

Web site tools can produce models suited for Web site tools can produce models suited for NeptuneNeptuneNeptune then maps these to actual codeNeptune then maps these to actual codeProducing a direct relationship between Producing a direct relationship between expert knowledge and an SOAexpert knowledge and an SOA


AcknowledgementsAcknowledgementsAcknowledgements

Sponsors and PartnersEPSRCUK NHS (Christies and Linda McCartney NHS trusts, EL Dental Services)IBM (PhD Fellowship) and WITSA (Egovernment)

The researchers and staff: www.cms.livjm.ac.uk/2nrichwww.cms.livjm.ac.uk/cloudwww.cms.livjm.ac.uk/taleb


That’s the end – so I’m off !


Neptune in Action: Clove ServicesNeptune in Action: Clove Services

autonomic computing meets complex information · pdf fileautonomic computing meets complex...

Documents