- from a Vision to Design and Implementation Peter...
Transcript of - from a Vision to Design and Implementation Peter...
![Page 1: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/1.jpg)
1
www.gridminer.org … Intelligent Grid Solutions
Knowledge Discovery Grids
Peter Brezany
Institute of Scientific Computing University of [email protected]
- from a Vision to Design and Implementation
Beijing, September 2005www.gridminer.org
Motivation
Business
Medicine
Scientificexperiments
SimulationsEarth observations
Data and dataexploration
cloud
Data and dataexploration
cloud
![Page 2: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/2.jpg)
2
Beijing, September 2005www.gridminer.org
Motivation
Data
Business understanding
Dataunderstanding
DataPreparation
Modeling
Evaluation
Deployment
CRISP-DM, SPSS
ServiceProvider
ServiceProvider
ServiceProvider
Data provider
Gri
dM
iner User
Virtual Organization
Beijing, September 2005www.gridminer.org
Outline
MotivationGrid – Evolution TrajectoryKnowledge Discovery in DatabasesGridMiner System Developed in Vienna
Scientific and Application DriversArchitectureWorkflow ManagementGraphical User InterfaceData Integration (Mediation) ConceptsServices for Data Mining and On-Line Analytical ProcessingPerformance Issues
![Page 3: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/3.jpg)
3
Beijing, September 2005www.gridminer.org
Outline (2) Towards Future Developments
Wisdom Web (Future Interconnection Environment)Current Grids and Next-Generation GridsSemantic Grids (GridMiner+)
ArchitectureWorkflow HierarchyOther Research Challenges
Conclusions GridMiner Demonstration
Beijing, September 2005www.gridminer.org
Media That Radically Influenced Society
Web
1500sPrinting Press
1840sPenny Post
1850sTelegraph
1920sTelephone
1930sRadio
1990s
1950sTV
20xxGrid
![Page 4: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/4.jpg)
4
Beijing, September 2005www.gridminer.org
Internet vs. Grid Computing
"The Internet is aboutgetting computers to talk together;
Grid computing is aboutgetting computers to work together."
Tom Hawk, IBM's general manager of Gridcomputing
Beijing, September 2005www.gridminer.org
Grid Computing Concept
The Grid – a new distributedcomputing infrastructure for science, engineering, and business.
The Grid consists of physicalresources (computers, disks, networks, files, sensors, laboratoryequipments, etc.) and “middleware“software that ensures the accessand the coordinated use of such resources.
It enables communities (“virtual organizations”) to share geographically distributed resources as they pursue common goals.
![Page 5: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/5.jpg)
5
Beijing, September 2005www.gridminer.org
Grid Development PhasesBasic Grid development (Globus 1) –
metacomputingData Grid (Globus 2, DataGrid of CERN,
etc.)Semantic Grid (myGrid)Open Grid Service Architecture (Globus 3,
OGSA-DAIS) & WSRF (Globus 4)Knowledge Discovery Grid (GridMiner and
work of others)Knowledge Grids & Web Intelligence (China
& Japan, GridMiner+)
Beijing, September 2005www.gridminer.org
Data Exploration Issues and theGridMiner Project in Vienna
![Page 6: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/6.jpg)
6
Beijing, September 2005www.gridminer.org
Knowledge Discovery Process in GridMiner
DataWarehouse
Knowledge
Cleaning andIntegration
Selection andTransformation
Data Mining
Evaluation andPresentation
OLAP
Online Analytical Mining
OLAP Queries
Data and functional resources can be geogra-phically distributed – focus on virtualization.
Beijing, September 2005www.gridminer.org
GridMiner (Goal) Architecture
GMMSMediation
GMPPSPreprocessing
GMDMSData Mining
GMPRSPresentation
GM DSCEDynamic Service Control
GMDTTransformation
GMOMSOLAM
GMISInformation
GMRBResource Broker
GridMiner Core
GMCMSOLAP / Cubes
GridMiner Base
GridMiner Workflow
Grid CoreServices
Security File and DatabaseAccess Service
ReplicaManagement
Grid Core
Grid Resources Data Sources
Fabric Basic GridServices
SMDSupport for Mobile Devices
GridMiner Mobility
![Page 7: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/7.jpg)
7
Beijing, September 2005www.gridminer.org
Working with GridMiner
GridMinerVisible GridPart
Invisible GridPart
Database and Gridadministrator End User
administrationcommands &
queries
data explorationworkflow spec. &
results
Beijing, September 2005www.gridminer.org
The Grid offers..
Theoretically, the Grid can have unlimited size (the number of data and computational resources) – support for scaling up
Questions:When is it necessary to mine huge databases, as opposed to mining
a sample of the data?Should not data mining algorithms be able to take advantage of all the data that is available?
Answers:Scaling up is desirable, because increasing the size of the trainingset often increases the accuracy of induced classification models.Determining how much data to use is difficult, because the smallestsufficient amount depends on factors not known a priori.Today‘s mining techniques can have problems when data setsexceed 100 megabytes.
![Page 8: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/8.jpg)
8
Beijing, September 2005www.gridminer.org
Optimizing predictive accuracyacc
ura
cy
sampled data size
100%
available data size
(qo,mo)
(qo,mo)
qi - data qualitymi - modeling method
(q0,m0)
Assumed
(qo,mo)
(qi,mi)
Beijing, September 2005www.gridminer.org
Pilot Application: Project EcoGRID
Waste
Air
Soil
Water
Emmision
Bio-diversity
Forests
DistributedData
DistributedData Mining
Flow Analysis
Geo-Statistic
Reporting
PopularPresen-tation
PredictionModels
DistributedApplications
…
Statistic
Common Ontology Author: Kathi Schleidt
![Page 9: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/9.jpg)
9
Beijing, September 2005www.gridminer.org
Management of TBI patients
Traumatic brain injuries (TBIs) typically result fromaccidents in which head strikes an object.The treatment of TBI patients is very resourceintensive.The trajectory of the TBI patients management:
Trauma eventFirst aidTransportation to hospitalAcute hospital careHome care
All the above phases are associated with data collectioninto databases – now managed by individual hospitals.
Usage of mobile communication
devices
Beijing, September 2005www.gridminer.org
Collaboration of GM-Services
GMPPSPreprocessing
GMDMSData Mining
GMDISIntegration
GMPRSPresentation
Data SourcesIntermediateResult 1
IntermediateResult 2(e.g. “flat table”)
IntermediateResult 3(e.g. PMML)
FinalResult
Simple Scenario:
![Page 10: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/10.jpg)
10
Beijing, September 2005www.gridminer.org
Collaboration (2)
GMDIS
GMPPS
GMPPS
GMPPS GMDMS GMPRS
GMPPS GMPPS
GMDMS
GMDMS
GMPRS
GMPRS
Complex Scenarios:
GMDMS GMPRS
GMDIS
GMPPS
GMPPS
GMCMS GMOMS GMPRSGMPPS
Beijing, September 2005www.gridminer.org
Workflow Models
Static Workflows Dynamic Workflows
![Page 11: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/11.jpg)
11
Beijing, September 2005www.gridminer.org
Workflows
DSCE
Service A Service B
Service C
Service D
DSCL
User
Dynamic Service Control Engine (DSCE)
processes the workflow according to DSCL
Dynamic Service Control Language (DSCL)
based on XMLeasy to use
DSCE Client
Beijing, September 2005www.gridminer.org
Workflow language (DSCL)
composition
dscl
sequence
parallel
invoke activityID=“act2.1” …
invoke activityID=“act2.2” …
createService activityID=“act1” …
sequence
variables
act1
act2.1
act2.2
…
User´s viewConversion
to XML
![Page 12: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/12.jpg)
12
Beijing, September 2005www.gridminer.org
Graphical User Interface – End-User Level
Beijing, September 2005www.gridminer.org
![Page 13: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/13.jpg)
13
Beijing, September 2005www.gridminer.org
Beijing, September 2005www.gridminer.org
Administration Level
![Page 14: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/14.jpg)
14
Beijing, September 2005www.gridminer.org
Beijing, September 2005www.gridminer.org
Grid Database Access
![Page 15: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/15.jpg)
15
Beijing, September 2005www.gridminer.org
Grid Database Access With OGSA-DAI
GDS gets a query via Perform Document
GDS Engine processspecified activities
GDS returns results
Beijing, September 2005www.gridminer.org
Grid Database Access With OGSA-DAI
![Page 16: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/16.jpg)
16
Beijing, September 2005www.gridminer.org
Grid Data Mediation Service
Heterogeneities:Name in A is „Alexander Wöhrer“Name in C has to be combined
Distribution:3 data sources
Example Scenario
Beijing, September 2005www.gridminer.org
Grid Data Mediation Service - Architecture
![Page 17: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/17.jpg)
17
Beijing, September 2005www.gridminer.org
OLAP (On-Line Analytical Processing)
Research Objectives: High- PerformanceGrid OLAP Services
Beijing, September 2005www.gridminer.org
Relational (ROLAP) vs. Multi- Dimensional Model (MOLAP)
Colour = {Red, Blue, White, Green}
Model = {Mini Van, Coupe, Sedan}
Two-Dimensional Model
![Page 18: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/18.jpg)
18
Beijing, September 2005www.gridminer.org
ROLAP vs. MOLAP
SELECT Model, Year, Color, SUM(sales) AS SalesFROM SalesWHERE Model IN {´Ford´, ´Chevy´} AND
Year BETWEEN 1990 AND 1992GROUP BY CUBE (Model, Year, Color);
Model Year Color Sales
Chevy 1990 Blue 87Chevy 1990 Red 5Chevy 1990 ALL 92Chevy ALL Blue 87Chevy ALL Red 5Chevy ALL ALL 92Ford 1990 Blue 99Ford 1990 Green 64Ford 1990 ALL 163Ford 1991 Blue 7Ford 1991 Red 8Ford 1991 ALL 15Ford ALL Blue 106Ford ALL Green 64Ford ALL Red 8All 1990 Blue 186All 1990 Gree 64ALL 1991 Blue 7ALL 1991 Red 8Ford ALL ALL 178ALL 1990 ALL 255ALL 1991 ALL 15ALL ALL Blue 193ALL ALL Green 64
Model Year Color Sales
Chevy 1990 Red 5Chevy 1990 Blue 87Ford 1990 Green 64Ford 1990 Blue 99Ford 1991 Red 8Ford 1991 Blue 7
⇒
Beijing, September 2005www.gridminer.org
ROLAP vs. MOLAP
Cross Tab Data Cube
![Page 19: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/19.jpg)
19
Beijing, September 2005www.gridminer.org
ROLAP vs. MOLAP
The ROLAP‘s summary records are storeddirectly int standard relational tables, withoutany need for data conversion. A complexanalytical query is cumbersome to express in SQL and it might not be efficient to execute.The array-based model, MOLAP (Multi-Dimensional OLAP), has the advantage thatnative arrays provide an immediate form of indexing for cube queries.
Beijing, September 2005www.gridminer.org
MOLAP
Each value of the attribute associated with thedimension determines a position of the dimension.The observed values, called measures are located at the intersections of the dimension positions. Theintersections are called cells and are populated withmeasures.Dimensional indexing: a unique integer value isassigned to each possible dimension position, and thecorresponding mapping information is stored in an appropriate index structure.
![Page 20: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/20.jpg)
20
Beijing, September 2005www.gridminer.org
Requirements
Operation on large data sets
Centralized OLAP Service (parallel computing power can be included)
Distributed OLAP service
Federation of autonomous distributed OLAP services
Beijing, September 2005www.gridminer.org
Development Strategy
Network
OLAP Engine
OE
OE
OE
OE
![Page 21: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/21.jpg)
21
Beijing, September 2005www.gridminer.org
Development Strategy (2)
Precondition: No open-source OLAP system availableDecision: development (in Java) fromscratchAdvantage: motivation for researchactivities addressing all facetsDisadvantage: a possible longimplementation curveFirst step: centralized sequential Grid OLAP service
Beijing, September 2005www.gridminer.org
Towards Centralized Service
GUIWorkflowEngine Mediator
OLAP
RD XMLD CSV
DSCL,OMML OMML XML
Data MiningEngine
PMML
PMML
![Page 22: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/22.jpg)
22
Beijing, September 2005www.gridminer.org
Distributed OLAP – Aggregation of Computeand Storage Resources
Tuple Stream
Beijing, September 2005www.gridminer.org
OLAP Service
Virtual Cube
Sub Cube
Sub Cube
Slave 1
Slave 3
Master
Data
Sub Cube
Slave 2
Indexes
Index Service
query
answerXML
![Page 23: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/23.jpg)
23
Beijing, September 2005www.gridminer.org
GridMiner Current Architecture
GUI
Knowledge Base
Service Configurators
Dynamic service composition engine (DSCE)
Data Base Access DM/OLAP services
Grid
Web
Use
r en
viro
nm
ent
DSCE Client
Beijing, September 2005www.gridminer.org
Towards an Open Service System
Clients(anywhere)
GUI (JWS)
Server(Vienna)
Cluster(Vienna)
Cluster(Linz)
Cluster(Innsbruck)
Workflow engine
WebApplications(Globus)
(Tomcat)
OLAP Master(Globus)
OLAP Slaves(Globus)
OGSA-DAI
Data mining services
(Tomcat)
(Globus)
![Page 24: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/24.jpg)
24
Beijing, September 2005www.gridminer.org
Implementation/Technology
Globus 3.2
OGSA/DAI version 5
GUI – Workflow constructions/Results visualization (JGraph, Java Web Start, Java server pages)Service Configurators (Java server pages)
Workflow management – DSCE Client (OGSA)
Knowledge base – Configurations (XML,OWL)
Data mediation service (OGSA/DAI)
Beijing, September 2005www.gridminer.org
From Data Mining Grid to Semantic Grid
![Page 25: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/25.jpg)
25
Beijing, September 2005www.gridminer.org
Towards the Next-Generation Web
The WWW is a milestone in the history of information sharing.A much more powerful infrastructure is neededto support e-Science.To overcome the Web‘s shortcomings, there isa vision of the next-generation Web, called theWisdom Web, also called Future Interconnection Environment. It is basedon the Web Intelligence concepts.
Beijing, September 2005www.gridminer.org
Topics Related to Web Intelligence
Diagram shows interrelated research topics from an intelligent Web-based, business perspective.
Ubuquitous computingand social intelligence
Web informationretrieval
Web miningand farming
Web informationmanagement
IntelligentWeb-based business
Knowledge networksand managementEmerging
Web technologyand Infrastructure
Intelligenthuman-Webinteraction
Web agents
Zhong, et al
![Page 26: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/26.jpg)
26
Beijing, September 2005www.gridminer.org
Towards the Next-Generation Web (2)
Focusing on two approaches:
Improving the existing Web, whichincludes Semantic Web, Web Services, and Web IntelligenceEstablishing a new collaborativecomputing perspective of the Web –the Grid Intelligence
Beijing, September 2005www.gridminer.org
Grid Intelligence Support for the Wisdom Web
Wisdom Web
Web Intelligence
SocialIntelligence
GridIntelligence
. . .
. . . . . .
. . .
Next-Generation
Grids
Based on Cheung & Liu [2005]
![Page 27: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/27.jpg)
27
Beijing, September 2005www.gridminer.org
Computational Grid
Data Grid
Data Minig Grid
Semantic Grid – 1st Generation
Current Grids
Next-GenerationGrid
Evolution ofthe Web
KnowledgeTechnologies
Evolution of HPCNMobileServices
Towards Next-Generation Grids
Beijing, September 2005www.gridminer.org
New EU IT Program
![Page 28: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/28.jpg)
28
Beijing, September 2005www.gridminer.org
Data Mining Grids
VEGA (Italy) – an environment running on top of Globus 2. No public demonstration of VEGA is known.Discovery Net (UK, project finished in 2004) – based on Globus 2.GridMiner (Austria, active project) – supportfor all phases of the knowledge discoveryprocess and OLAP. Based on OGSA and OGSA-DAI.
Beijing, September 2005www.gridminer.org
Goals
The aim is not only knowledge discovery and sharing among scientists, but also
sustainable knowledge creation and scientific evolution
Reusing
Knowledge life cycle
Discovery
Sharing
Processing
![Page 29: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/29.jpg)
29
Beijing, September 2005www.gridminer.org
Semantic Grids – 1st GenerationMost projects mainly focus on “Knowledgefor the Grid“ (using knowledge technologies to describe availability of Grid services, how they arediscovered, invoked, ...).
This approach is also partially followed in GridMinerOur new focus on “Knowledge on the Grid“ New framework planned: GridMiner+Research challenges:
Problem Solver Markup LanguagesAccess to and integration of knowledge basesHigh-performance distributed reasoningDynamic multi-level workflowsAutonomic features
Beijing, September 2005www.gridminer.org
Working with GridMiner+
GridMiner+Visible Grid
PartInvisible Grid
Part
Ontology, database, and Grid administrator User
administrationcommands &
queries
data explorationworkflow specifications,
results
problemspecifications,
solutions
![Page 30: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/30.jpg)
30
Beijing, September 2005www.gridminer.org
GridMiner+ - Architecture
Auto
nom
ic S
upport
Intelligent interface
Knowledge management infrastructure
Knowledge Provider
Problem Solver
Generic Grid Services
Globus Toolkit
Data mining infrastructure
GridMiner
Beijing, September 2005www.gridminer.org
Knowledge Base
Metadata
Rules
Facts
Ontologies
Models
InferenceEngine
Rules and FactsGenerator
![Page 31: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/31.jpg)
31
Beijing, September 2005www.gridminer.org
Transform request queries
Validate requestwith ontologies
Process request
User Problem solver Knowledge provider(GridMiner)
Search KBs
Processresults
Transform request to KDD task
KDDWorkflow
Generate rules
Processresults
Find data miningactivity
Apply rules on domain ontology
Generate outputs
KDD
Problem Solving WorkflowsStatic and Dynamic Features
understandable?
yes
no
Demand
Response
Response
PSML – Problem Solver Markup Language
DB AccessWorkflow
Beijing, September 2005www.gridminer.org
PSMLs
SWRL Semantic Web Rule Language (W3C): combination of OWL with RuleMLβ-PSML (Su et al.): combination of OWL withHorn clauses; it can represent multi-argumentrelations.Query example (Prolog notation):
?- hasHypertensionRisk(patient#12756)
hasHypertensionRisk, its values, etc. aredefined in the appropriate domain ontology
![Page 32: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/32.jpg)
32
Beijing, September 2005www.gridminer.org
Rules
The rulehasHypertensionRisk(patient) :-
systolicGreaterThan(patient,140),diastolicGreaterThan(patient,90),periodInYearsGreaterThan(patient,2).
If the rule is not in the KBs, the systemstarts data mining and tries to derive therule automatically.
Beijing, September 2005www.gridminer.org
Distributed Reasoning
Needed for solving problem in a large scale Web and Grid environment.Appropriate PSML features need to be investigated.Two architectures can be considered:
KB KB KB
IE
GN GN
GN
GN
GN – Grid NodeKB – Knowledge BaseIE – Inference Engine
IE KB
IE KB IE KB
GN
GN
GN
![Page 33: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/33.jpg)
33
Beijing, September 2005www.gridminer.org
Autonomic Computing Support
Providing data mining and knowledgemanagement services transparently
Relieving professionals from activities thatare of a system nature
Allowing them to concentrate their effortson their area activities
Offering computing and communivationservices quickly, reliably and securely.
Beijing, September 2005www.gridminer.org
Conclusions
In the near future, the development of intelligent applications will be based on intelligent interconnection environments (IIEs)
IIEs will be supported by the Web and Grid Intelligence
Grid Intelligencerealized by Semantic Gridsimportant components
Data Mining Gridsknowledge basesefficient reasoning mechanisms
![Page 34: - from a Vision to Design and Implementation Peter Brezanyhomepage.univie.ac.at/peter.brezany/teach/scientD... · Institute of Scientific Computing University of Vienna brezany@par.univie.ac.at](https://reader034.fdocuments.in/reader034/viewer/2022050412/5f88e010e729e91504693c20/html5/thumbnails/34.jpg)
34
Beijing, September 2005www.gridminer.org
Demo Presentation (Video) Follows