The Great OLAP Debate! TM1, PowerPlay & DMRs April 29, 2011.
-
Upload
derrick-page -
Category
Documents
-
view
213 -
download
0
Transcript of The Great OLAP Debate! TM1, PowerPlay & DMRs April 29, 2011.
The Great OLAP Debate!
TM1, PowerPlay & DMRs
April 29, 2011
Panel Presenters
Michael Langton
Scott Luck-Baker
Mike Roberts
Pedro Mendoza
Panel Debate Format
Each one of the panelists will present evidence that their approach is the best way to handle OLAP reporting
Your job as a participant is to ask questions to challenge each OLAP approach …
Goals For Session IBM provides several options for OLAP reporting
Does one size fit all?
We will review each technology:
Description Product Background Key Functionality
Business Use Case “Sweet Spot”
Usage Notes, Design and Deployment Considerations
IBM TM1
IBM TM1 - Overview
Developed in the late 80’s as a backend for spreadsheets
Multi-dimensional database designed to simplify complex spreadsheets and separate the data from the formulas
TM1 cubes are essentially collections of business hierarchies (dimensions); numeric and text data can be stored at the intersections of every dimension element
TM1 cubes sit in RAM so that data consolidation and formulas (cube rules) are performed in “real-time”
TM1 clients include Excel, TM1Web, Contributor (for workflow), Executive Viewer, and Cognos BI
TM1 Web
TM1 Contributor
Executive Viewer
IBM TM1 - Sweet Spot
TM1 is designed for the WRITEBACK of numeric and text data
It is ultimately flexible and models can be built from a variety of data and meta-data sources to hold almost any type of data
TM1 includes a rule language for writing complex formulas into your model; rules are evaluated in real-time for instant feedback
Non-technical users can perform administrative/modeling tasks via wizards, drag & drop actions, or using customizable buttons through the web
Users can slice & dice cube views, and drill through to further levels of detail
IBM TM1 - Usage Cases
Replacing Excel as a planning tool
Slicing & dicing aggregated data
Comparing apples to apples
Processes that require manual entry
Processes that require real-time feedback/calculations
Processes that require workflow/security
What-if analysis / Driver-based planning
New ways of rolling up your data
Anywhere non-technical users need to build reports, add/remove elements, launch imports/exports
IBM Cognos PowerPlay
IBM Cognos PowerPlay- Overview
Originally developed by Cognos in 1989
PowerPlay Transformer is used to define OLAP cube structures and building static “cubes” for analysis or reporting, usually on a scheduled basis
PowerPlay Cubes contain summarized data organized into dimensions and measures, can be built from very large datasets and are highly optimized for data retrieval
PowerPlay Cubes can be viewed via the web (Analysis Studio, Query Studio, Report Studio, C10 Biz Insight/Advanced) or via a full client (PowerPlay Client, CAFÉ Excel)
IBM Cognos PowerPlay- Sweet Spot Ideal where users have large datasets that require flexible
summarization and reporting options, as opposed to a list of canned reports
Transformer provides advanced multi-dimensional model support and varied data sources (via Framework Manager), incremental refresh options, alternate drill paths, automatic category counts, time-based and volume-based partitioning strategies
Non-technical users can explore data through simple click and drag operations and can gain insight through functions such as rank, sort, nesting and calculations
Users can drill from cube-to-cube or cube-to-database
IBM Cognos PowerPlay- Usage Cases
Great for sales, marketing and financial analysis
Users have large datasets, possibly in an existing database or across multiple sources
Users or IT want “self serve” analysis capabilities
Users want “zero footprint”
Users have no interest in budgeting or planning
Users don’t need real time reporting
Users don’t need to report on “non-dimensional data” elements
Warning: Prone to User Misuse (esp. in S7)
DMR Framework Designs
DMR Framework Designs - Overview
Introduced in Cognos 8
Uses Framework Manager to model Relational Data to appear “like a cube”
Model can be used in Analysis Studio and Report Studio Express
Define Regular Dimensions Consists of one or
more user-defined hierarchies
Each hierarchy consists of
levels
keys
captions
attributes
Edit DMR in the Dimension Map View, create, or modify:
regular or measure dimensions
hierarchies or levels
scope relationships
What the Authors SeeDimension
Hierarchy
Level
Member
Child members
Report StudioData Tree
DMR Framework Designs - Sweet Spot
You do not need another application to build cubes
No need to wait while cube is being built
Data changes in the underlying tables are immediately available
Complex security rules can be created in one place (Framework Manager)
Define multiple Hierarchies for a Dimension
Define as many member attributes as you want
DMR Framework Designs - Usage Cases
Implement Drill Up/Down in reports without cubes
Analysis of Real-time data or data that would take too long to build into a cube
Models that have complex business rules that would be difficult to implement in a cube
Solutions where security is defined at the database level
DMR – Deployment Considerations
Aggregations are not stored in a cube. They are calculated from detail every time a report or analysis is run.
Performance is dependent on good hardware and design
Databases must be optimized to maximize performance
It may be necessary to employ a form of database vendor materialization to improve performance
DMR packages are usually built on top of existing FM models and are deployed the same way.
DMR – Design Considerations
Design for Performance
Physical data should be in star schemas to minimize complex joins
Create summary tables to avoid aggregating on the fly
Model for high level analyses and rely on drill-through reports to give detail
DMR works best with small narrow dimensions rather than large wide dimensions
Build on top of a good well-designed relational Framework.
Build mandatory filters into your model to ensure that end users do not accidentally retrieve excessively large data sets
Panel Debate