Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Watson Research Center
Real-time Data Integration Real-time Data Integration for for Business Activity Business Activity
MonitoringMonitoring
Josef SchieferIBM Watson Research
Workshop in Adaptive Business Operations02-25-2003
>> A Container Based Approach>> A Container Based Approach
Watson Research Center
Gaining Insight into the Business
Tech
nolo
gy E
volu
tion Data Warehousing
Business Process Warehousing
Real-Time Business Monitoring
Adaptive Decision Support Systems
Technology Evolution Technology Evolution forforGaining Insight into the BusinessGaining Insight into the Business
Watson Research Center
Wor
kflo
w M
anag
emen
t Sys
tem Automatic Feedback
Input: Event DataOperation: Automatic Comparison, EvaluationResponse: Adjustment of available quantities(e.g. workload-based staff assignment, changing priorities)
Strategic Management & Control
Input: High-level Quality ParametersOperation: Manual Comparison, AlarmsResponse: Adjustment of process and environment
Operative Management & ControlInput: Process-specific Quality ParametersOperation: Manual Comparison, AlarmsResponse: Adjustment of available quantities (e.g. schedulingadditional staff members, changing inventory levels) Filter
FilterOperative System
Management System
Instant (real-time) response time for decision making can be guaranteed (seconds); theperiod of time during which the decision has an effect is very short. Decisions are made
very frequently and instantly when the events are raised.
Time horizon: months – years. A decision is made often only once.Historical data is very critical. Includes high-level business metrics.
Leve
ls o
f Le
vels
of
Res
pon
sive
nes
sR
espo
nsi
ven
ess
Watson Research Center
Task 1 Task 2 Task 3 Task 4 Task 5 ProcessOutcomes
Customer
Sub Process 1Sub Process 2
M3
M4 M4
M3
M4 M4 M4
M3 M1M1
Internal
ExternalM2
Manager ObjectivesM3
- Reduce service costs- Reduce order processing time- Improve service quality- Optimize resource assignments
Workgroup ObjectivesM4
- Reduce defects/waste/delays- Improve customer rating- Automate processing steps- Early error detection
Vice Pres./Director ObjectivesM2
- Shorten delivery and supply time- Increase Return on Investment- Increase delivery performance
Executive ObjectivesM1
- Fully exploit customer potential- Improve profitability- Customer satisfaction
Process Monitoring LevelsProcess Monitoring Levels
Watson Research Center
Data Integration ChallengesData Integration Challengesfor BAMfor BAM
Data Propagation- Near real-time- Continual- Complex ETL processing- Many, but small data extracts
Adding Business Process Context- Context consists of
* Process Description* Organizational Context* Business Data* Causal Context Data
- Context information must becollected from several data sources
- Business Metrics on various abstraction levels
Configurability / Manageability- Configuration Changes in Runtime- Deployment in existing DWH environments- Reflection of data propagation status
Evaluation of Business Metrics- Evaluation of SLAs- Response Mechanisms- Notifications
Solution Management- Solution Templates- Reusable, pluggable ETL components- Adaptive Architecture
Watson Research Center
J2EE ETL EnvironmentJ2EE ETL Environment
Pro
prie
tary
Con
nect
ors
EJB Container
J2EE
ETL
Env
ironm
entJava Services
JAAS, JNDI, JMS,JAXP, JCE, JTA,
JAF, ...
ETL Container
Res
ourc
e A
dapt
ers
Evaluators
RangeEvaluators
CustomEvaluators
CompareEvaluators
Event Adapters
JMS EventAdapter
JCA EventAdapter
MQ SeriesEvent Adapter
JCA EventAdapter...
ETLets
Event DrivenETLets
ScheduledETLets
ExceptionETLets
Session Beans
ConversionEJBs
AssemblyEJBs
ParsingEJBs
CleansingEJBs...
CMPEntity Beans
BMPEntity BeansE
AI/B
PI
Con
nect
ors
Dat
abas
eC
onne
ctor
sJ2
EEC
onne
ctor
sJM
SC
onne
ctor
s
Sour
ce S
yste
ms
ERP
...
Database
External Sources
WFMS
Legacy Systems
Message DrivenEJBs
Watson Research Center
ETL Container - PurposeETL Container - Purpose
Extracting raw event data from various source systems
Calculation of business metricsPersisting event data and business metrics
Coordination of ETL processing/transformation (--> processing flow)
Evaluation of business metricsAllows flexible (re-)configuration/deployment of ETL components
Separates extraction, transformation, and evaluation logic(--> pluggable components)
Provides a configurable, scalable and high-performance ETL environment for a large number of events
Watson Research Center
Batch vs. ContinuousBatch vs. ContinuousData IntegrationData Integration
OperationalSource
OperationalSource
OperationalSource
OperationalSource
WFMS
Batch Data IntegrationLoading
TransformationExtraction
EnterpriseData Warehouse
ProcessData Store
ProcessWarehouse
Business ProcessMonitoring
ETL Container
EventAdapters
ETLets Evaluators
ProcessInformation
Factory
WorkflowEvents
Response
Notifications
Workflow Metrics
Watson Research Center
Event Adapters: Dequeue events and transform them into a standard event formatETLets process the standardized events and calculate the business metricsEvaluators evaluate the calculated business metrics
Real-time Data Integration Real-time Data Integration with an ETL Containerwith an ETL Container
(Event Processing, Calculation & Evaluation of Business Metrics)(Event Processing, Calculation & Evaluation of Business Metrics)
Cleansing(Parsing, Correction,
Standardization,Completion)
Matching Consolidation,Transformation
MetricCalculations
ProcessWarehouse
Message Queue
J2EE Connector
JDBC Connector
EventAdapters ETLetsStandardized
XML Events
ProcessData Store
EvaluatorsMetric
Evaluation
Rule EnginesETL Container
...
Watson Research Center
Real-Time Event ProcessingReal-Time Event Processing
Managed straight-through ETL processing without intermediary storage!!
ETL Container
Eve
nt A
dapt
er Thread 1
Thread 2
Thread 3
EventHandler
ETLetService
ETLet 1
ETLet 2
ETLet 3
...
EvaluationService
Evaluator 1
Evaluator 2
Evaluator 3
MetricHandler
MetricDispatcher
Event 1
Event 3
Event 2
Eve
nt D
ispa
tche
r -- Workflow Event Processing --E
vent
Ada
pter
Eve
nt A
dapt
er
ProcessData Store
ProcessWarehouse
...
Watson Research Center
BPSM ArchitectureBPSM Architecture
Policy Managem
ent
EventAdapters
ETLets Evaluators
Pro
cess
Eve
nts ETL Container
ProcessWarehouse
ProcessData Store
Message Bus
BusinessMetrics
Message Bus
ProcessInformation
Factory
BI Agent Layer
Dashboard Facade
IntegrationLayer
ReactiveAgent
DeliberateAgents
ReflectiveAgents
Policies
SensingAgents
ResponseAgents
WFMS
Process 2
Process 3
Process 1 BusinessApplications
Service BeansUser BeansManagement Widgets
Dashboard
Presentation Layer
Watson Research Center
ETL Container - StrengthsETL Container - Strengths
Lightweight, near real-time ETL processing (Threads)
Support of complex transformations / processing
Evaluation capabilities of calculated business metrics
Usage of existing J2EE infrastructure
Container managed, optimized, reusable, configurable ETL components
Clean separation of Extraction logic (EventAdapters)
Transformation logic (ETLets)
Evaluation logic (Evaluators)
Event handling is separated from the database
Event-driven actions linked to the near real-time data integration
Development methodology very similar to Web Applications (with Servlets)
Watson Research Center
When NOT to use Near Real-Time Data Integration:
Near real-time data propagation is NOT a requirement
Existing data replication solution are sufficient
Very large data volumes (--> bulk loading)
When NOT to use an ETL container:Transformations can be done with SQL or even not necessary
Configurability is provided by ETL tools
No Java development skills available
When NOT to use Near Real-Time When NOT to use Near Real-Time Data Integration / ETL ContainerData Integration / ETL Container
Watson Research Center
Future Work: Future Work: Adaptive BAM SolutionsAdaptive BAM Solutions
Process 2
Process 3
Process 1
BPEL
BPEL
BPEL
WFMS ETL Container
Even
t Ada
pter Thread 1
Thread 2
Thread 3
EventHandler
ETLetService
ETLet 1
ETLet 2
ETLet 3
...
EvaluationService
Evaluator 1
Evaluator 2
Evaluator 3
MetricHandler
MetricDispatcher
Event 1
Event 3
Event 2
Eve
nt D
ispa
tche
r -- Processing of Workflow Events --
Even
t Ada
pter
Eve
nt A
dapt
er
ProcessData Store
ProcessWarehouse
PIF Builder
SchemaDefinition/
Composition
SchemaGeneration
SchemaDefinitions
Configuration
WorkflowEvents
ETLConfiguration
Process Start Time
Customer
Coverage
Process
activity_time_keypolicy_keycustomer_keyagent_keycoverage_keycoverage_item_keyprocess_keyprocess_outcome_keyamountsum_waiting_timeprocessing_timeprocess_runtimeprocess_runtime_devprocess_state_times…
Policy
Agent
Covered Item
Process Outcome
Process Compl. TimeProcess Start Time
Customer
Coverage
Process
activity_time_keypolicy_keycustomer_keyagent_keycoverage_keycoverage_item_keyprocess_keyprocess_outcome_keyamountsum_waiting_timeprocessing_timeprocess_runtimeprocess_runtime_devprocess_state_times…
Policy
Agent
Covered Item
Process Outcome
Process Compl. Time
1 2 3
...
BPCL
Watson Research Center
Thank You!Thank You!
Questions?Questions?
Watson Research Center
Backup Slides
Backup
Backup Slides
Watson Research Center
Integrated data foundation for the monitoring & analysis of business processes. The PIF adds the process context to existing DWH environments
Enables process-oriented view of all business data
Provides new process key performance indicators
Near real-time integration of workflow events
Process Information Factory (PIF)Process Information Factory (PIF)with Container Managed Data Integrationwith Container Managed Data Integration
Workflow Data
Business Data Integrated process-orienteddata foundation for a
process-oriented monitoring/analysis
Watson Research Center
Transaction Processing vs.Transaction Processing vs.Informational ProcessingInformational Processing
Level 1 - Transactional Processing
OrderProcessing
Level 2 - Informational Processing
Fulfillment Billing Payment
...
TR-75
Miller
Singer
List
Risher
TR-75
Miller
Singer
List
Risher
531 539 652 683 867531 539 652 683 867
Jun.May.
Apr.Mar.Feb.Jan.
Jun.May.
Apr.Mar.Feb.Jan.
Sale
sA
ssis
tant
Sale
sM
anag
er VI
DEO
all
TraditionalSale
InternetSale
1. quarter
2. quarterHJ1
1. quarter
2. quarterHJ1
Participant
i me
P r o c e s s
Jones
Sale
Uni
t
2727
531Jan.
Miller
Singer
List
Rahl
T
O r g a n i z a t i o n
Real-tim
e InformationR
eal-t
ime
Dat
a In
tegr
atio
n
ProcessWarehouse
ProcessData Store
EnterpriseData Warehouse
Watson Research Center
Comparison:Comparison:DWH, ODS, PDS, PWHDWH, ODS, PDS, PWH
Criteria:Basic orientation:
Business view:Process information:
Time references:Access:
Aggregation level:Integration level:
Accessibility:
Legend: Data managed by transaction systemsData managed by operational data storesData managed by traditional data warehouse systemsData managed by process data storesData managed by process warehouses
transactionlocal
no informationactual
read/writedetailedisolated
real-time
subject-orientedenterprise-widefull informationactual + historyread-onlyaggregatedintegrateddelayed
Watson Research Center
Metric Handler( calls the metric evaluators)
Binding between ETLets andMetric Evaluators are defined in the deployment descriptor
Metric Handler( calls the metric evaluators)
Binding between ETLets andMetric Evaluators are defined in the deployment descriptor
Implementation: ETLets, EvaluatorsImplementation: ETLets, Evaluators