2011 central tug tawfik tbio
TRANSCRIPT
-
8/2/2019 2011 Central TUG Tawfik TBIO
1/26
2011 Spring TUG
Sam TawfikProduct Marketing Manager
[email protected] @teradata_sam
Teradata Business IntelligenceOptimizer
mailto:[email protected]:[email protected] -
8/2/2019 2011 Central TUG Tawfik TBIO
2/26
Agenda
OLAP MOLAP/ROLAP Teradata Aggregate Designer
Teradata OLAP Connector Demo
Implementation Considerations
-
8/2/2019 2011 Central TUG Tawfik TBIO
3/26
OLAP Environments
Marketing Customer ServiceFinance/Planning
Physical Cubes(MOLAP)
Physical Cubes(MOLAP)
Physical Cubes(MOLAP)
Multiple BIEnvironments
Cubes everywhere
Data replication
Long ETL jobs
Limited scalability
Limited data sets
Mid-tier server costs
-
8/2/2019 2011 Central TUG Tawfik TBIO
4/26
Convert to ROLAP EnvironmentsEliminate Physical Cubes
Marketing Customer ServiceFinance/Planning
Physical Cubes(MOLAP)
Physical Cubes(MOLAP)
Physical Cubes(MOLAP)
-
8/2/2019 2011 Central TUG Tawfik TBIO
5/26
ROLAP Solution
Marketing Customer ServiceFinance/Planning
BI Server BI Server BI Server
Physical Cubes(MOLAP)
Physical Cubes(MOLAP)
Physical Cubes(MOLAP)
Multiple BIEnvironments
Single View
Eliminate delays
DW scalability
Rich detailed data
Reduce/Eliminate need
for Mid-tier server
-
8/2/2019 2011 Central TUG Tawfik TBIO
6/26
ROLAP Solution
Solution Convert MOLAP to ROLAP with Teradata Database
> Let Teradata host and analyze the data
> Use existing Teradata Indexes to optimize response time
> Use Teradata Aggregate Join Index (AJI) the Teradata
built-in OLAP optimization capabilities
Benefits High performance analytics powered by Teradata BI solutions benefit from the following:
> Increased breadth/depth of business information> Eliminate data latency> Eliminate/reduce need for the mid-tier server> Avoid mid-tier scalability and performance limitations
-
8/2/2019 2011 Central TUG Tawfik TBIO
7/26
OLAP Implementations Attributes
Attribute MOLAP HOLAP ROLAP
Flexibility Application Specific Application Specific Application Neutral
Query
Performance
Sub-secondsSub-seconds for cube;few seconds for detail
Few seconds
Breadth ofAnalytics
Limited to Cube onserver
Almost unlimited Almost Unlimited
Depth ofAnalytics
Limited to Cube onserver and drillthrough
Drill-down for detailedanalysis
Detailed analysis
Data Freshness Varies depending oncube refresh
Combination of cubedata and current
Current data
Cost ofImplementation
High Moderate Low
-
8/2/2019 2011 Central TUG Tawfik TBIO
8/26
Customer Example
38 dimensions and 24 measures with twoyears of history
> Added 39th dimension (Wire Center)
MOLAP solution couldnt handle thebusiness change
Implemented ROLAP solution using AJIs
> History expanded from 24 to 38 months
> # of dimensions expanded from 38 to 50
> # of measures expanded from 15 to 21
Results
> Maintenance: 13 hoursto3 minutes
> Cube size: 22.4 GB to Detail: Month to Daily
ROLAP Implementation
0
50
100
150
200
250
300
5 Canned 5 Canned with Wire
Center
6 Interactive 6 Interactive with Wire
Center
MOLAP
ROLAP with AJIs
Response Comparison:
-
8/2/2019 2011 Central TUG Tawfik TBIO
9/26
Teradata Aggregate Designer
-
8/2/2019 2011 Central TUG Tawfik TBIO
10/26
AJIs IntroducedV2R3.0.2
Sparse JI/AJI, Single AMP, Group-AMPV2R5.0
RIV2R5.1
PPIV2R6.2
M-PPI12
PGB, Cost-based re-writeOUTER JOIN, COUNT,
CUBE, ROLLUP, GROUPING SET,Subqueries, Spool derived table
13
Teradata Aggregate Join Index Timeline
-
8/2/2019 2011 Central TUG Tawfik TBIO
11/26
Aggregate Join IndexPre-aggregated Results and Calculations
For frequently needed roll-up summaries> Sales results, orders, shipments, etc.
> Bypass join, sort, sum at run time
> Applies for lower level aggregations (many transactions)
Good for
> Dashboards, portal
> Interactive analytics
pre-aggregatedresults
Fast responsetime from using
aggregated results
-
8/2/2019 2011 Central TUG Tawfik TBIO
12/26
ROLAP Implementation with Teradata
1. Identify the target solution start with a small cube2. Capture the cube schema definition
3. Validate the BI schema and physical data model are suitable for ROLAP
4. Perform data quality checks
5. Create the semantic data model
6. Select the target database for the ROLAP solution
7. Select the aggregation levels for each dimension
8. Create the necessary base and broad AJI(s)
9. Review the Aggregate Join Index DDLs and modify as necessary
10. Validate/Publish the AJIs on the target database
11. Perform one data loading job from source to the fact table
12. Test the AJIs to make sure they are being used by the Optimizer13. Collect statistics
14. Tune the AJIs as the BI solution evolves
15. Gradually add more features to the cube or create new cubes
-
8/2/2019 2011 Central TUG Tawfik TBIO
13/26
Description> Desktop visual administrative tool to automate designing,
building, and managing AJIs
> Natively import cube definitions from IBM Cognos, Microsoft SQLServer Analysis Services, or Teradata Schema Workbench
Release 13.10 Enhancements
> Native support for IBM CognosFramework Manager(8.3, 8.4, and 10)
> Ease of use and functionalityenhancements
> Enhanced database and schema validations
> Detect and alert Schema changes
> Support AJI enhancement in 13.10
Teradata Aggregate Designer
-
8/2/2019 2011 Central TUG Tawfik TBIO
14/26
Teradata Aggregate Designer Features
Database Validation> Validates target database schema
for AJI requirements
> Recommends possible databaseschema changes
AJI Advisor
> Uses Teradata-specific heuristics torecommend the optimal AJI design
AJI Creation Services
> Provides AJI storage cost estimates
> User can manually modify AJI levels
> Validates the SQL and creates
Teradata DDL
Supports PI and PPI definitions
-
8/2/2019 2011 Central TUG Tawfik TBIO
15/26
Teradata Aggregate Designer automatesdesigning, building, and managing AJIs for ROLAPsolutions
Semantic Layer
Office SharePoint Server
SQL ServerAnalysis Services
Office
Performance
PointServer
SQLServer
Reporting
Services
Teradata .NET Data Provider
Excel
PivotTables
TeradataAggregate Designer
AJIsAJIsAJIs
Workflow
1. DBA leverages the Teradata Aggregate Designer
tool to consume cube schema from SQL ServerAnalysis Services
2. DBA creates/modifies necessary AJIs usingTeradata Aggregate Designer
Optimized SQL
Microsoft BI with Teradata ROLAPLeveraging Teradata Aggregate Designer
-
8/2/2019 2011 Central TUG Tawfik TBIO
16/26
Portal
MetaData
Scorecard
Cognos Server
Reporting
Dashboard
Workflow
1. BI Admin creates and manages OLAP schema in
the Cognos Framework Manager.2. Teradata Aggregate Designer consumes model.xml.
3. Teradata Aggregate Designer validates database issuitable for Aggregate Join Indexes. Recommendsand creates AJIs.
Teradata Aggregate Designer automates thedesign and deployment of AJIs for BI Reporting,Dashboards, ROLAP, and MOLAP solutions
IBM Cognos with Teradata ROLAPLeveraging Teradata Aggregate Designer
TeradataAggregate Designer
AJIsAJIsAJIs
Optimized SQL
-
8/2/2019 2011 Central TUG Tawfik TBIO
17/26
Teradata Aggregate Designer Demo
-
8/2/2019 2011 Central TUG Tawfik TBIO
18/26
Teradata OLAP ConnectorExcel to Teradata Connectivity
Teradata OLAPConnector
MDX
SQL
1. Open a new Excel workbook
2. Establish data connection to Teradata Database via Teradata OLAP Connector
3. Select the project (catalog), and then select the cube
4. Use Excels PivotTables to navigate/view the data Teradata OLAP Connector convertsPivotTable requests into optimized Teradata SQL
-
8/2/2019 2011 Central TUG Tawfik TBIO
19/26
Teradata OLAP Connector Demo
-
8/2/2019 2011 Central TUG Tawfik TBIO
20/26
Implementation Considerations
-
8/2/2019 2011 Central TUG Tawfik TBIO
21/26
Implementation Approach
Moving from MOLAP to ROLAP requires planningand changes to the BI application
Customers should balance the work to be doneto migrate from MOLAP against theadded benefits of ROLAP
Recommendations
1. Start with a ROLAP PoC to validate the approach
Typically the PoC is completed in a short period of time and with asmall budget so make sure you select an appropriate cube to model
The PoC should focus on the ROLAP solution itself and not the BIchanges
Consider using Excel PivoTables to demonstrate the navigation andreporting of the ROLAP implementation
2. Once the ROLAP PoC has been completed then plan for widerimplementation
-
8/2/2019 2011 Central TUG Tawfik TBIO
22/26
Data Loading Considerations
Teradata Multiload does not support AJIs, following arethe different load scenarios to consider:
> Multiload into normalized tables, and then build AJIs fromthe normalized tables
> Teradata Tpump into normalized tables
> Teradata Fastload into staging tables, and then insert selectinto the base tables
-
8/2/2019 2011 Central TUG Tawfik TBIO
23/26
Teradata Database Considerations (1 of 2)
We recommend you start with star and snowflake models All primary and foreign keys are defined as not nullable
All primary and foreign keys are not compressible.
All dimension table primary keys are defined as unique
Ensure all primary and foreign keys are on ID not Name or
Description columns. This will result in a smaller AJI whichmeans faster access.
Ensure all measures in the Fact table are set with Floatdatatype. Otherwise, overflow may occur on some largecalculated measures
Teradata Release 12 or later
-
8/2/2019 2011 Central TUG Tawfik TBIO
24/26
Teradata Database Considerations (2 of 2)
Recommended Fact Table design is a 'wide' (i.e. columns for eachdimension and measure)
Single level dimensions need supporting reference/lookup/dimensiontable for optimal performance
Collect statistics on all primary key/foreign key relationship columns
Implement Referential Integrity (RI) on the primary key/ foreign keycolumns. RI can be defined with the check (a.k.a., hard RI) or nocheck option (a.k.a., soft RI).
> ORGS: Business Unit -> Division -> Area -> Sales Center
> ALTER TABLE FACT ADD foreign key (SALES, CENTER_ID) references withno check option SALE_CENTER (SALES_CENTER_ID)
To address the lowest levels in your analytics (i.e. your FACT) thatare not in your AJI, consider the following:
> Secondary Indexes
> Partition Primary Index
-
8/2/2019 2011 Central TUG Tawfik TBIO
25/26
Performance and Capacity Assessmentsand Services
Data Collection and Reporting Collect
and analyze key performance metrics Workload Management Deliver maximum
throughput of mixed workloads based
on query mix and business priorities
Application Performance Identify target
applications for performance improvements andtune those applications
Performance Assessment Quick hit service desinged to analyze andidentify opportunities for performance improvements
Capacity Management Understanding growth requirements andfuture needs based on application growth, performance requirements,
and user sophistication Viewpoint Advisor Service Best practices implementation of
Viewpoint Administration and Portlet use
Performance Management Foundation Foundational ServiceDesigned around Making the Customers Performance Team selfsufficient in all these areas
CapacityManagement
DataCollection
-
8/2/2019 2011 Central TUG Tawfik TBIO
26/26