mastering intelligent clouds engineering intelligent data processing services in the cloud sergiy...

20
Mastering Intelligent Mastering Intelligent Clouds Clouds Engineering Intelligent Data Processing Engineering Intelligent Data Processing Services in the Cloud Services in the Cloud Sergiy Nikitin, Sergiy Nikitin, Industrial Ontologies Industrial Ontologies Group, Group, University of Jyväskylä, University of Jyväskylä, Finland Finland Presented at Presented at ICINCO 2010 conference ICINCO 2010 conference Funchal, Madeira Funchal, Madeira

Upload: emmett-folkes

Post on 22-Dec-2015

221 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Mastering Intelligent CloudsMastering Intelligent Clouds

Engineering Intelligent Data Processing Engineering Intelligent Data Processing Services in the CloudServices in the Cloud

Sergiy Nikitin,Sergiy Nikitin,

Industrial Ontologies Group,Industrial Ontologies Group,

University of Jyväskylä, FinlandUniversity of Jyväskylä, Finland

Presented atPresented at

ICINCO 2010 conferenceICINCO 2010 conference

Funchal, MadeiraFunchal, Madeira

Page 2: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

ContentsContentsContentsContents

• Background on Cloud ComputingBackground on Cloud Computing

• Extending cloud computing stackExtending cloud computing stack

• UBIWARE platformUBIWARE platform

• Data Mining services in the CloudData Mining services in the Cloud

• ConclusionsConclusions

Page 3: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Cloud Computing: already on the marketCloud Computing: already on the marketCloud Computing: already on the marketCloud Computing: already on the market

• SalesForce.com (SFDC)SalesForce.com (SFDC)• NetSuiteNetSuite• OracleOracle• IBMIBM• MicrosoftMicrosoft• Amazon EC2Amazon EC2• GoogleGoogle• etc.etc.• (for a complete survey see Rimal et al., 2009)(for a complete survey see Rimal et al., 2009)

Page 4: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Cloud Computing stackCloud Computing stackCloud Computing stackCloud Computing stack

Cloud computing

stack

PaaS

SaaS

IaaS

Hardware configuration

Virtualization Machine

OS-virtualization

Raw data storage and network

Structured storage (e.g. databases)

Solution stack (Java, PHP, Python, .NET)

Services (Payment, Identity, Search)

Application (business logic)

Application as a Service

What add-value can we offer to the PaaS level?

Page 5: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Autonomic ComputingAutonomic ComputingAutonomic ComputingAutonomic Computing

• A vision introduced by IBM in 2003 (Kephart et al.)A vision introduced by IBM in 2003 (Kephart et al.) software components get a certain degree of self-software components get a certain degree of self-

awarenessawareness self-manageable components, able to “run themselves”self-manageable components, able to “run themselves”

• Why?Why? To decrease the overall complexity of large systemsTo decrease the overall complexity of large systems To avoid a “nightmare of ubiquitous computing” – an To avoid a “nightmare of ubiquitous computing” – an

unprecedented level of complexity of information unprecedented level of complexity of information systems due to:systems due to:

• drastic growth of data volumes in information systems drastic growth of data volumes in information systems • heterogeneity of ubiquitous components, standards, data heterogeneity of ubiquitous components, standards, data

formats, etc.formats, etc.

Page 6: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Intelligence as a Service in the cloudIntelligence as a Service in the cloudIntelligence as a Service in the cloudIntelligence as a Service in the cloud

Agent-driven service APIConfiguration management

Solution stack

Domain models

Data adaptation

Intelligent servicesPaaS

Structured storage (e.g. databases)

Solution stack (Java, PHP, Python, .NET)

Services (Payment, Identity, Search)

UBIWARE

• Smoothly integrate with the infrastructureSmoothly integrate with the infrastructure• Build stack-independent solutionsBuild stack-independent solutions• Automate reconfiguration of the solutionsAutomate reconfiguration of the solutions

Page 7: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

UBIWARE platformUBIWARE platformUBIWARE platformUBIWARE platform

.cla

ss

BlackboardBlackboard

Role Script

RA

BR

AB

RA

BR

AB

RA

BR

AB

RA

BR

AB

Beliefs storage

UBIWARE AgentUBIWARE Agent

Pool of Atomic Pool of Atomic BehavioursBehaviours

S-APL S-APL repositoryrepository

S-APLS-APLData

Page 8: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Cloud Platform Provider

Virtual machine

SW Platform

Customer applications and services

Ext

en

de

d A

PI

PCA PMA

API extension: OS perspectiveAPI extension: OS perspectiveAPI extension: OS perspectiveAPI extension: OS perspective

PCA – Personal Customer Agent

PMA – Platform Management Agent

Page 9: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Data Adaptation as a ServiceData Adaptation as a ServiceData Adaptation as a ServiceData Adaptation as a Service

Cloud Platform Provider

Virtual machine

SW Platform

Customer applications and services

Ext

en

de

d A

PI

PCAPMA

PCA – Personal Customer Agent

PMA – Platform Management Agent

Adapter Adapter AgentAgent

FilesFiles

Data Service

DB/KBDB/KB

Page 10: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Cloud Platform Provider

Virtual machine

SW Platform

Customer applications A

PI

Virtual machine

API

Service execution

environmentPCA

PMA

Platform-driven service execution in the cloudPlatform-driven service execution in the cloudPlatform-driven service execution in the cloudPlatform-driven service execution in the cloud

PCA – Personal Customer Agent

PMA – Platform Management Agent

Page 11: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Agent-driven PaaS API extensionAgent-driven PaaS API extensionAgent-driven PaaS API extensionAgent-driven PaaS API extension

Agent-driven flexible intelligent service API

Agent-driven Adapters

Agent-driven intelligent services

Smart cloud stack

Smart Ontology

Standards & compatibility

System configuration and policies

Domain models

Failure-prone maintenance

Stack control and updates

Embedded and remote services

Service mobility

Configurable model

Proactive self-management

Smart data source connectivity

Configurable data transformation

Proactive adapter management

Use

r ap

plic

atio

ns in

clo

udU

ser

appl

icat

ions

in c

loud

Page 12: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Intelligent services: PaaS API extensionIntelligent services: PaaS API extensionIntelligent services: PaaS API extensionIntelligent services: PaaS API extension

Agent-driven flexible intelligent service API

Agent-driven intelligent services

Service mobility

Configurable model

Proactive self-management

Use

r ap

plic

atio

ns in

clo

udU

ser

appl

icat

ions

in c

loud

Page 13: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Agent-driven data mining servicesAgent-driven data mining servicesAgent-driven data mining servicesAgent-driven data mining services

ModelInput Output

Vector DM model DM result

Agent serviceAgent service

Data mining applications are capabilitiesData mining applications are capabilities Agents can wrap them as servicesAgents can wrap them as services PMML language - a standard for DM-model representationsPMML language - a standard for DM-model representations

Data Mining Group. PMML version 4.0. URL http://www.dmg.org/pmml-v4-0.html

Page 14: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Header

Model development environment information

Version and timestamp

PMML model

Data dictionary

Definition of: variable types, valid, invalid and missing values

Data Transformations

Data aggregation and function calls

Normalization, mapping and discretization

Model

Description and model specific attributes

Mining schema

Definition of: usage type, outlier and missing value treatment and replacement

Targets

Definition of model architecture/parameters

Score post-processing - scaling

PMML*: data mining model descriptionsPMML*: data mining model descriptionsPMML*: data mining model descriptionsPMML*: data mining model descriptions

PMML* - Predictive Model Markup Language (www.dmg.org/pmml-v3-0.html)

Page 15: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Data mining service typesData mining service typesData mining service typesData mining service types

ModelInput Output

Vector to be classified:alarm message:

V1={0.785, High, node_23}

Paper machine alarms classifier neural

network model (M1)

Vector class of V1 is:“Urgent Alarm”

according to model M1

Fixed model serviceFixed model service

Model player serviceModel player service

Model construction serviceModel construction service

ModelInputs Outputs

Set up a modelM1

Paper machine alarms classifier neural

network model (M1)

Model M1 assigned

Vector to be classified:alarm message:

V1={0.785, High, node_23}

Model player

Vector class of V1 is:“Urgent Alarm”

according to model M1

ModelInput Output

Learning samples and the desired model settings Model M1 parametersModel constructor

Page 16: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

A use case for data mining service stackA use case for data mining service stackA use case for data mining service stackA use case for data mining service stack

ModelInput Output

Pattern of learning data to be collected:

?V={?p1, ?p2, ?p3}

Distributed query planning and

execution

A set of learning samples (vectors)

1

Learning samples and the desired model settings Model M1 parametersModel constructor

Set up a modelM1

Paper machine alarms classifier neural network

model (M1)

Model M1 assigned

Vector to be classified:alarm message:

V1={0.785, High, node_23}

Model player

Vector class of V1 is:“Urgent Alarm”

according to model M1

2

3

4

A “Web of Intelligence” case:A “Web of Intelligence” case:

Page 17: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Data Mining services in UBIWAREData Mining services in UBIWAREData Mining services in UBIWAREData Mining services in UBIWARE

Data Mining service

Model construction service

Computational service

Fixed model service

Model player service

Ontology constructionOntology construction

Core DM service ontology

Core DM service ontologyData mining domain

Data mining domain

Problem domainProblem domain

Mining method

Supervised Learning

Unsupervised learning

Clustering kNNNeural networks

Industry

Process Industry Electrical Engineering

Power networks Power plantPaper industry

Page 18: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

UBIWARE in cloud computing stackUBIWARE in cloud computing stackUBIWARE in cloud computing stackUBIWARE in cloud computing stack

Cloud computing stack

Technologies in cloud

Platform as a

service

Applications and Software as a Service

Infrastructure as a service

Hardware configuration

Virtualization Machine

OS-virtualization

Raw data storage and network

Structured storage (e.g. databases)

Solution stack (Java, PHP, Python, .NET)

Services (Payment, Identity, Search)

Application (business logic)

Application as a Service

DM model for paper industry

Example application

DM model wrapped as a service for paper industry

Cross-domain Middleware components

Componentization & Servicing

Connectors, Adapters

RABs, Scripts

UBIWARE for control and management in cloud

Semantic Business Scenarios

Domain model (Ontology) & components

Domain-specific components as services

Cross-layer configuration & management mechanisms

Agent-driven service API

Data Mining service player

Page 19: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

ConclusionsConclusionsConclusionsConclusions

• Web intelligence as a cloud serviceWeb intelligence as a cloud service

• Ubiware is a cross-cutting management and Ubiware is a cross-cutting management and configuration glueconfiguration glue

• Advanced data adaptation mechanisms as cloud Advanced data adaptation mechanisms as cloud servicesservices A competitive advantage for cloud providersA competitive advantage for cloud providers Seamless data integration for service consumption and Seamless data integration for service consumption and

provisioningprovisioning

• Autonomous agents as a Service (A4S)Autonomous agents as a Service (A4S) Supply any resource with the “autonomous manager”Supply any resource with the “autonomous manager”

Page 20: Mastering Intelligent Clouds Engineering Intelligent Data Processing Services in the Cloud Sergiy Nikitin, Industrial Ontologies Group, University of Jyväskylä,

Thank you!Thank you!Thank you!Thank you!