Presenter: Khalid Saleem

Enabling autonomic Grid Applications: Enabling autonomic Grid Applications: Requirements, Models and Requirements, Models and

InfrastructureInfrastructure

Authors:Authors:M. Parashar, Z. Li, H. Liu, V. Matossian and C. M. Parashar, Z. Li, H. Liu, V. Matossian and C.

SchmidtSchmidt

Presenter: Khalid SaleemPresenter: Khalid Saleem

IntroductionIntroduction Investigate challenges and Investigate challenges and

requirements of programming Grid requirements of programming Grid ApplicationsApplications

Self-Management ApplicationsSelf-Management Applications Introduce Project AutomateIntroduce Project Automate

Investigates Autonomic solutions to deal withInvestigates Autonomic solutions to deal with ComplexityComplexity DynamismDynamism HeterogeneityHeterogeneity Uncertainty Uncertainty

Grid Computing – Challenges Grid Computing – Challenges & Requirements (1)& Requirements (1)

Goal:Goal: Applications utilizing intellectual and physical Applications utilizing intellectual and physical

resources across different disciplines and resources across different disciplines and organizations organizations

Providing more effective and efficient Providing more effective and efficient solutions to important Scientific, Engineering, solutions to important Scientific, Engineering, Business and Government problemsBusiness and Government problems

Built on Seamless and secure mechanisms forBuilt on Seamless and secure mechanisms for DiscoveryDiscovery AccessAccess InteractionsInteractions


Key characteristics of Grid Key characteristics of Grid Execution Environment and Execution Environment and Applications areApplications are

HeterogeneityHeterogeneity DynamismDynamism UncertaintyUncertainty SecuritySecurity


Goals attainment requiresGoals attainment requires Implementation ModelsImplementation Models Conceptual ModelsConceptual Models

Implementation Models addressImplementation Models address Virtualization of organizations (GRIDs) Virtualization of organizations (GRIDs) Creation and Management of virtual organizationsCreation and Management of virtual organizations Instantiation of a virtual machine as the execution Instantiation of a virtual machine as the execution

environment of an applicationenvironment of an application Conceptual Models defineConceptual Models define

Abstract machines supporting programming models Abstract machines supporting programming models and systems enabling application developmentand systems enabling application development


Requirements for Grid Programming Requirements for Grid Programming SystemsSystems

Specify applications that can detect and dynamically Specify applications that can detect and dynamically respond during execution to the changes inrespond during execution to the changes in

state of execution environmentstate of execution environment state and requirements of the application state and requirements of the application

Grid Applications should Grid Applications should Constitute of discrete, self managing componentsConstitute of discrete, self managing components Provide separate specifications for functional, non-Provide separate specifications for functional, non-

functional, interaction and coordination behaviorsfunctional, interaction and coordination behaviors Provide implementation independent interface Provide implementation independent interface

definitions for above specifications definitions for above specifications

Requirements for Self-Requirements for Self-managing Applications on the managing Applications on the

GridGrid Self-managing applications on the Grid Self-managing applications on the Grid requirerequire

Static application requirementsStatic application requirements System and Application behaviors should be relaxedSystem and Application behaviors should be relaxed

Sensitivity and Ability of elements and applications to Sensitivity and Ability of elements and applications to adapt to theadapt to the

Dynamic state of the system; andDynamic state of the system; and Changing requirements of the applicationChanging requirements of the application

Essential common knowledge be expressed Essential common knowledge be expressed semantically (i.e. Ontology) rather than names semantically (i.e. Ontology) rather than names addresses and identifiersaddresses and identifiers

Core Middleware services (e.g. Discovery) driven by Core Middleware services (e.g. Discovery) driven by semantic knowledge semantic knowledge

AutoMateAutoMate An attempt to address the above challenges and An attempt to address the above challenges and

requirementsrequirements

AutoMate AutoMate Investigates autonomic solutions to Investigates autonomic solutions to

address challenges posed to Grid address challenges posed to Grid ComputingComputing

Based on Biological Systems Based on Biological Systems Development of Conceptual and Development of Conceptual and

Implementation models/architecturesImplementation models/architectures Programming Models, Framework and Programming Models, Framework and

Middleware ServicesMiddleware Services Definition of autonomic elementsDefinition of autonomic elements Development of autonomic applicationsDevelopment of autonomic applications Policy, Content and Context driven application Policy, Content and Context driven application

definition, management and execution definition, management and execution

Components of AutoMateComponents of AutoMate

Accord: A Programming Accord: A Programming Framework for Autonomic Framework for Autonomic

ApplicationsApplications Extends existing distributed programming Extends existing distributed programming

systemssystems RealizesRealizes

Separation of computations from coordination and Separation of computations from coordination and interactionsinteractions

Separation of non-functional behaviors (e.g. Resource Separation of non-functional behaviors (e.g. Resource requirements, performance) from functional behaviorsrequirements, performance) from functional behaviors

Separation of policy and mechanismSeparation of policy and mechanism Accord Programming Model key componentsAccord Programming Model key components

Autonomic ElementsAutonomic Elements Accord RulesAccord Rules Autonomic CompositionAutonomic Composition

Autonomic Elements (1)Autonomic Elements (1) Autonomic ElementAutonomic Element

Extends programming elementsExtends programming elements Objects, Components and ServicesObjects, Components and Services

Defines self-contained Modular software unit Defines self-contained Modular software unit withwith

Specified interfacesSpecified interfaces Explicit context dependenciesExplicit context dependencies

Encapsulates Encapsulates RulesRules ConstraintsConstraints MechanismsMechanisms

Can dynamically interact with other elements Can dynamically interact with other elements and the systemand the system

Autonomic ElementsAutonomic Elements (2)(2)

Autonomic Element Autonomic Element

Autonomic Elements (3)Autonomic Elements (3) Autonomic Element defined by 3 portsAutonomic Element defined by 3 ports

Functional PortFunctional Port Defines a set of functional behaviors provided and Defines a set of functional behaviors provided and

used by the elementused by the element Control PortControl Port

Defines a set of sensors/actuators and the Defines a set of sensors/actuators and the corresponding access constraintscorresponding access constraints

Sensors: Interfaces providing info about the elementsSensors: Interfaces providing info about the elements Actuators: Interfaces for modifying state of the Actuators: Interfaces for modifying state of the

elementselements Operational PortOperational Port

Defines the interfaces to formulate and manage rules Defines the interfaces to formulate and manage rules that govern the runtime behavior of the elements that govern the runtime behavior of the elements and their interactions.and their interactions.

Autonomic Elements (4)Autonomic Elements (4) Element ManagerElement Manager

Manages element executionManages element execution Monitors state of the element and its Monitors state of the element and its

contextcontext Controls execution rulesControls execution rules Cooperates with other element managersCooperates with other element managers

Accord Rules (1)Accord Rules (1) Rules in the formRules in the form

IF condition THEN actionIF condition THEN action Condition: Logical combination of sensors, Condition: Logical combination of sensors,

elements, function interfaces and eventselements, function interfaces and events Action: Sequence of invocations of Action: Sequence of invocations of

element and/or system actuators and element and/or system actuators and other interfacesother interfaces

Priority based mechanism to Priority based mechanism to resolve conflictsresolve conflicts

Accord Rules (2)Accord Rules (2) Rule ClassesRule Classes

Behavioral RulesBehavioral Rules Controls runtime functional behaviors of an Controls runtime functional behaviors of an

autonomic elementautonomic element Executed by the element manager within a Executed by the element manager within a

single elementsingle element Interaction RulesInteraction Rules

Controls interactions and coordination of Controls interactions and coordination of autonomic elementsautonomic elements

Element managers collaborate to provide Element managers collaborate to provide coordinated execution coordinated execution

Autonomic Composition (1)Autonomic Composition (1) Enables relationship between elements to Enables relationship between elements to

be established and modified at runtimebe established and modified at runtime Dynamic composition consists of Dynamic composition consists of

composition plan and executioncomposition plan and execution Plans created at runtimePlans created at runtime

Based on dynamically defined objectives, policies Based on dynamically defined objectives, policies and the context and content of applications and and the context and content of applications and systemssystems

Plan execution involvesPlan execution involves Discovering elements, configuring them and Discovering elements, configuring them and

defining interaction relationships and mechanismsdefining interaction relationships and mechanisms

Autonomic Composition (2)Autonomic Composition (2) Composition Plans Composition Plans

Generated using Accord Composition Engine (ACE)Generated using Accord Composition Engine (ACE) Expressed in XMLExpressed in XML

Element Discovery usesElement Discovery uses Meteor content-based MiddlewareMeteor content-based Middleware Squid Discovery ServiceSquid Discovery Service

Plan executionPlan execution P2P control network of element managers and P2P control network of element managers and

agents within Rudder agents within Rudder Library of rule setsLibrary of rule sets

Common control and communication relationships Common control and communication relationships between elementsbetween elements

Runtime Negotiation ProtocolsRuntime Negotiation Protocols Address runtime conflicts and conflicting decisionsAddress runtime conflicts and conflicting decisions

Accord Implementation Accord Implementation IssuesIssues

Accord allows interaction & coordination Accord allows interaction & coordination behaviors to be managed at runtime using behaviors to be managed at runtime using rulesrules

Deploying and executing rulesDeploying and executing rules Impacts performanceImpacts performance Increases robustness of applications; and Increases robustness of applications; and Ability to manage dynamismAbility to manage dynamism

Runtime changes to interaction and Runtime changes to interaction and coordination behaviors relatively infrequent coordination behaviors relatively infrequent and have small overheadsand have small overheads

Time spent in establishing and modifying Time spent in establishing and modifying interactions is small as compared to interactions is small as compared to computation timescomputation times

References [35, 36]References [35, 36]

Rudder Coordination Rudder Coordination FrameworkFramework

Scalable coordination middlewareScalable coordination middleware Supports self managing applications in Supports self managing applications in

decentralized distributed environmentdecentralized distributed environment Consists of two key componentsConsists of two key components

COMETCOMET Enables flexible and scalable coordination among Enables flexible and scalable coordination among

agents and autonomic elementsagents and autonomic elements Agent FrameworkAgent Framework

Composed of software agents, agent interaction Composed of software agents, agent interaction and negotiation protocolsand negotiation protocols

COMET (1)COMET (1) Provides a global virtual shared Provides a global virtual shared

coordination spacecoordination space Accessible by all peer agents, independent of Accessible by all peer agents, independent of

physical location of tuples or identifiers of the hostphysical location of tuples or identifiers of the host Build on an associative messaging substrateBuild on an associative messaging substrate

Implements a distributed hash tableImplements a distributed hash table Index space generated from semantic information Index space generated from semantic information

space (ontology)space (ontology) Can be transient and dynamically constructedCan be transient and dynamically constructed

Provides an associative communication Provides an associative communication abstractionabstraction

Maps virtual information space to a dynamic set of Maps virtual information space to a dynamic set of available peer nodesavailable peer nodes

Maintains content localityMaintains content locality

COMET (2)COMET (2) Provides a coordination abstraction layerProvides a coordination abstraction layer

Extends traditional data-driven model with event Extends traditional data-driven model with event based reactivity to changes inbased reactivity to changes in

System stateSystem state Data access operationsData access operations

Defines a Reactive tuple abstraction consisting ofDefines a Reactive tuple abstraction consisting of ConditionCondition that associates that associates ReactionReaction to events to events A A GuardGuard specifying how and when the specifying how and when the ReactionReaction will be will be

executedexecuted Provides Linda-like primitives Provides Linda-like primitives

http://en.wikipedia.org/wiki/Linda_(coordination_language)http://en.wikipedia.org/wiki/Linda_(coordination_language)

Agent Framework (1)Agent Framework (1) Composed of Dynamic Network of Software Composed of Dynamic Network of Software

Agents existing at different levelsAgents existing at different levels Agents are Agents are

Processing unitsProcessing units Perform actions based on dynamically defined rulesPerform actions based on dynamically defined rules

AgentsAgents Monitor element statesMonitor element states Manage element behaviors and dependenciesManage element behaviors and dependencies Coordinate element interactionsCoordinate element interactions Cooperate to manage overall system/application behaviorCooperate to manage overall system/application behavior

Agents use profiles toAgents use profiles to Identify and describe elementsIdentify and describe elements Interact with elementsInteract with elements Control elementsControl elements

Agent Framework (2)Agent Framework (2) Defines a set of protocols forDefines a set of protocols for

Agent coordination Agent coordination Application/system managementApplication/system management

Examples of protocols includeExamples of protocols include Discovery protocolsDiscovery protocols Control ProtocolsControl Protocols

Meteor: A content-based Meteor: A content-based MiddlewareMiddleware

Scalable content-based Middleware Scalable content-based Middleware infrastructureinfrastructure

Consists ofConsists of Self organizing content overlaySelf organizing content overlay Content based routing engine and discovery service Content based routing engine and discovery service

(Squid)(Squid) Associative Rendezvous Messaging Substrate Associative Rendezvous Messaging Substrate

(ARMS)(ARMS)

Meteor Content OverlayMeteor Content Overlay Composed of Rendezvous Peer nodes Composed of Rendezvous Peer nodes

(RP)(RP) RP nodes can join or leave the network at anytimeRP nodes can join or leave the network at anytime

Provides a single operation Provides a single operation Lookup (identifier)Lookup (identifier)

Locates the peer node where the content should be Locates the peer node where the content should be stored or fetched stored or fetched

SquidSquid Content based routing engine and Content based routing engine and

decentralized discovery servicedecentralized discovery service SupportsSupports

Flexible content-based routing Flexible content-based routing Complex queries containing partial keywords, Complex queries containing partial keywords,

wildcards and rangeswildcards and ranges Keywords can be common words or globally defined Keywords can be common words or globally defined

attributes based on ontologies or taxonomiesattributes based on ontologies or taxonomies Uses Hilbert Space Filling Curve (SFC)Uses Hilbert Space Filling Curve (SFC)

Maps multidimensional information space to peer Maps multidimensional information space to peer identifier spaceidentifier space

Associative Rendezvous Associative Rendezvous Messaging Substrate (ARMS)Messaging Substrate (ARMS)

Implements Associative Rendezvous (AR) Implements Associative Rendezvous (AR) interaction paradigminteraction paradigm

Content-based decoupled interactions with Content-based decoupled interactions with programmable reactive behaviorsprogrammable reactive behaviors

Allows for asynchronous interactions Allows for asynchronous interactions among senders and receiversamong senders and receivers

Extends conventional name/identifier Extends conventional name/identifier based rendezvous bybased rendezvous by

Using flexible combination of keywords, wildcards Using flexible combination of keywords, wildcards and ranges from the semantic information space and ranges from the semantic information space instead of identifiersinstead of identifiers

Enabling reactive behaviors at rendezvous points Enabling reactive behaviors at rendezvous points encapsulated within messages, thus increasing encapsulated within messages, thus increasing flexibility and enabling multiple interaction flexibility and enabling multiple interaction semantics (broadcast, multicast)semantics (broadcast, multicast)

Autonomic Grid Applications

Autonomic Oil Reservoir Autonomic Oil Reservoir Optimization (1)Optimization (1)

ProblemsProblems Selection of appropriate optimization algorithmsSelection of appropriate optimization algorithms Runtime configuration and invocation of these algorithmsRuntime configuration and invocation of these algorithms Dynamic optimization of reservoirDynamic optimization of reservoir

Autonomic application based on AutoMate consists of Autonomic application based on AutoMate consists of 1.1. Sophisticated reservoir simulation componentsSophisticated reservoir simulation components

Encapsulates complex mathematical models Encapsulates complex mathematical models Execution on GridExecution on Grid

2.2. Grid Services providing secure and coordinated access to Grid Services providing secure and coordinated access to resourcesresources

3.3. Distributed data archivesDistributed data archives Stores historical, experimental and observed dataStores historical, experimental and observed data

4.4. Sensors embedded in the instrumented oilfieldSensors embedded in the instrumented oilfield Provides real time data about the current state of oil fieldProvides real time data about the current state of oil field

5.5. External services providing data relevant to optimization of oil External services providing data relevant to optimization of oil production or the economic profitsproduction or the economic profits Optimization based on VFSA and SPSA algorithms Optimization based on VFSA and SPSA algorithms

6.6. Actions of scientists, engineers in field, labs and management Actions of scientists, engineers in field, labs and management officesoffices

Autonomic Oil Reservoir Autonomic Oil Reservoir Optimization (2)Optimization (2)

Peers automaticallyPeers automatically Detect sub-optimal production behaviors at runtimeDetect sub-optimal production behaviors at runtime Adjust interactions to correct sub-optimal productionAdjust interactions to correct sub-optimal production

Uses policies toUses policies to Discover, select, configure and invoke appropriate Discover, select, configure and invoke appropriate

optimization services to determine optimal well optimization services to determine optimal well location location

Autonomic Forest Fire Autonomic Forest Fire Management Simulation (1)Management Simulation (1)

Predicts the speed, direction and intensity of fire Predicts the speed, direction and intensity of fire frontfront

Uses dynamic environment and vegetation conditionsUses dynamic environment and vegetation conditions Consists ofConsists of

Data Space Manager (DSM)Data Space Manager (DSM) Partitions the forest represented by 2D data space into sub spaces Partitions the forest represented by 2D data space into sub spaces

based on information from CRMbased on information from CRM Computational Resource Manager (CRM) Computational Resource Manager (CRM)

Provides system resources information to DSMProvides system resources information to DSM RothermelRothermel

Simulates in parallel the fire spread in each sub spaceSimulates in parallel the fire spread in each sub space Uses current wind direction and intensity info simulated by Wind Uses current wind direction and intensity info simulated by Wind

ModelModel Wind ModelWind Model

Provides wind direction and intensity simulationsProvides wind direction and intensity simulations GUIGUI

Allows experts to interact with the above elementsAllows experts to interact with the above elements


Implementation of Accord Port typesImplementation of Accord Port types


Component Addition and Change in Interaction Relationship Example Component Addition and Change in Interaction Relationship Example

ConclusionConclusion AutoMateAutoMate

Implementation architecture and conceptual models Implementation architecture and conceptual models that enable the development and execution of self-that enable the development and execution of self-managing Grid applicationsmanaging Grid applications

Agnostic Question 1Agnostic Question 1 How does AutoMate deal with the heterogeneity among How does AutoMate deal with the heterogeneity among

all the elements of a grid? How does AutoMate all the elements of a grid? How does AutoMate integrate so many heterogeneous autonomic integrate so many heterogeneous autonomic components to work together?components to work together?

By separating the policy from mechanism. Policies in By separating the policy from mechanism. Policies in the form of rules are used to orchestrate a repertoire of the form of rules are used to orchestrate a repertoire of mechanisms mechanisms to achieve context-aware adaptive to achieve context-aware adaptive runtime computational behaviors and coordination and runtime computational behaviors and coordination and interaction relationships based on functional, interaction relationships based on functional, performance, and QoS requirements thus responding to performance, and QoS requirements thus responding to heterogeneity and dynamics. Uses Dynamic heterogeneity and dynamics. Uses Dynamic composition to enable relationships between elements composition to enable relationships between elements (via element managers) to be established and modified (via element managers) to be established and modified at runtimeat runtime

Agnostic Question 2Agnostic Question 2 Security and trust is one of the key issues in grid Security and trust is one of the key issues in grid

environments, and the maintenance of users and environments, and the maintenance of users and permissions may become a bottleneck. Does permissions may become a bottleneck. Does AutoMate provide a specific autonomic solution AutoMate provide a specific autonomic solution for a self-maintenance of users and permissions for a self-maintenance of users and permissions across administrative boundaries?across administrative boundaries?

AutoMate does not explicitly provide a AutoMate does not explicitly provide a methodology for self maintenance of users and methodology for self maintenance of users and permissions across administrative boundaries. permissions across administrative boundaries. However, SESAME that is embedded into However, SESAME that is embedded into AutoMate allows for setting up security policies AutoMate allows for setting up security policies and permissions for avoiding fraud and intrusion. and permissions for avoiding fraud and intrusion.

Agnostic Question 3Agnostic Question 3 The Accord framework incorporates practical human The Accord framework incorporates practical human

knowledge in the form of behavioral rules, but with knowledge in the form of behavioral rules, but with the vastness of communication paradigms (e.g., RPC, the vastness of communication paradigms (e.g., RPC, RMI, publish/subscribe), how does Accord ensure the RMI, publish/subscribe), how does Accord ensure the correctness of these rules?correctness of these rules?

““To ensure rule correctness, we design a rule To ensure rule correctness, we design a rule template for each communication paradigm (e.g., template for each communication paradigm (e.g., RPC/RMI, publish/subscribe) and coordination pattern RPC/RMI, publish/subscribe) and coordination pattern (e.g., conditional branch, loop, sequence, parallel (e.g., conditional branch, loop, sequence, parallel execution). A rule template is instantiated via filling execution). A rule template is instantiated via filling in parameters, such as the names of interaction in parameters, such as the names of interaction parties and the exchanging data.” parties and the exchanging data.” http://http://www.caip.rutgers.edu/~marialiu/Projects/Accord/index.htmwww.caip.rutgers.edu/~marialiu/Projects/Accord/index.htm

http://www.caip.rutgers.edu/~marialiu/Projects/Accord/index.htm

http://www.caip.rutgers.edu/~marialiu/Projects/Accord/index.htm

Agnostic Question 4Agnostic Question 4 AutoMate utilizes a peer-to-peer middleware to AutoMate utilizes a peer-to-peer middleware to

support and enable the autonomic interactions support and enable the autonomic interactions among components. Can you think of a better among components. Can you think of a better architecture to achieve this goal?architecture to achieve this goal?

In this scenario using peer-to-peer would be a In this scenario using peer-to-peer would be a better option as AutoMate is based upon better option as AutoMate is based upon autonomic agent based control networks. Element autonomic agent based control networks. Element managers execute rules to establish control and managers execute rules to establish control and communication relationships among these communication relationships among these elements in a decentralized and parallel manner.elements in a decentralized and parallel manner.

Agnostic Question 5Agnostic Question 5 Does Accord target a specific programming Does Accord target a specific programming

language?language?

Nothing particular. However, the prototype Nothing particular. However, the prototype implementation was done using C++ and MPIimplementation was done using C++ and MPI

Agnostic Question 6Agnostic Question 6 Can you extend on how AutoMate lets Globus-Can you extend on how AutoMate lets Globus-

enabled grids to achieve an autonomic enabled grids to achieve an autonomic management? What elements of the AutoMate management? What elements of the AutoMate architecture communicate with Globus architecture communicate with Globus elements?elements?

AutoMate makes use of its Accord autonomic AutoMate makes use of its Accord autonomic elements and rule set, the Rudder Agent elements and rule set, the Rudder Agent Framework and COMET to provide dynamic Framework and COMET to provide dynamic flexible and scaleable coordination among the flexible and scaleable coordination among the peer nodes along with the use of conflict peer nodes along with the use of conflict resolution and negotiation protocols. No specific resolution and negotiation protocols. No specific information mapping the components of information mapping the components of AutoMate to that of Globus has been provided.AutoMate to that of Globus has been provided.

Agnostic Question 7Agnostic Question 7 Are you aware of other projects with the same Are you aware of other projects with the same

goals than AutoMate?goals than AutoMate?

Grid MP by United Devices Grid MP by United Devices http://www.ud.com/products/gridmp_fabs.phphttp://www.ud.com/products/gridmp_fabs.php

Optimal Grid Optimal Grid

http://www-128.ibm.com/developerworks/library/ghttp://www-128.ibm.com/developerworks/library/gr-opgrid/r-opgrid/

GridARM for Structured Adaptive Mesh GridARM for Structured Adaptive Mesh Refinement (SAMR) application Refinement (SAMR) application

OGSA to some extentOGSA to some extent

http://www.ud.com/products/gridmp_fabs.php

http://www-128.ibm.com/developerworks/library/gr-opgrid/

http://www-128.ibm.com/developerworks/library/gr-opgrid/

Agnostic Question 8Agnostic Question 8 On a previous presentation of the same authors, On a previous presentation of the same authors,

we discussed that their so-called peer-to-peer we discussed that their so-called peer-to-peer architecture was in fact Client-Server. Do you architecture was in fact Client-Server. Do you think it is correct that they somewhat reuse this think it is correct that they somewhat reuse this architecture for the Meteor and COMET architecture for the Meteor and COMET components?components?

The previous presentation focused on PAWN The previous presentation focused on PAWN where as the authors made use of Meteor in the where as the authors made use of Meteor in the current framework. Meteor is based on JXTA which current framework. Meteor is based on JXTA which is a general purpose P2P framework. Hence, it is a general purpose P2P framework. Hence, it seems correct to use Meteor.seems correct to use Meteor.

Documents

Presenter: Khalid Saleem