If you can't read please download the document
Upload
alex-meadows
View
6.911
Download
0
Embed Size (px)
Citation preview
Mondrian and OLAP
Frontends
RTP Pentaho User GroupQ1 2012 Meetup
Pentaho Overview
BI Server (Frontend for tools)
Report Designer (canned report designer)
Mondrian (Schema workbench, aggregate designer)
Data Integrator
Other ad-hoc tools (reporting)
Weka (predictive analytics)
Enterprise Extras
Analyzer
Interactive Reporting
Dashboard Designer
Data Integration Scheduler
Support
OLAP?
On-line Analytical Processing
Designed for Analytics, not transactions
ROLAP (Mondrian)
Relational OLAP
MOLAP (Palo)
Multidimensional OLAP
ROLAP
Benefits
Data is stored in a Kimball-style star schemaUsable by all other tools (reporting, dashboards, etc.)
Cube is stored in memory
Cons
Performance while cube is being cached
Performance depending on backend database
MOLAP
Benefits
Data stored in multidimensional format
Usually highly compressed
Cons
Potentially long processing times to handle permutations
Higher cardinality (dimensions with millions of records) increases processing
Mondrian Development Life-cycle
ROLAP Optimizations
Columnar data stores
Built for huge datasets in a conformed dimension format
Highly compresses and scales
Examples: LucidDB, Infobright, InfiniDB
Mondrian Specific Optimizations
Aggregate Designer
Performs cost/benefit analysis on all permutations of data
Builds SQL queries that can be loaded into ETL or plugins (like with LucidDB) and run at set times
Cons
Have to refresh aggregate data as new data comes in will get stale otherwise!
Time to refresh is dependent on data set
MDX
MultiDimensional Expressions
SQL for OLAP
Open standard developed by Microsoft
MDX
Source: http://sqlblogcasts.com/blogs/drjohn/archive/2008/09/27/mdx-and-sql-combining-relational-and-multi-dimensional-data-into-one-query-result-set.aspx