SQL: Enough business intelligence time for administration intelligence
Download as PPTX, PDF0 likes516 views
The document discusses three core methods for applying business intelligence (BI) tools to administrative data: reporting, modeling, and mining. Each method serves different scenarios, with reporting being suitable for known fields and views, modeling allowing for multiple report types from a shared data model, and mining useful when the important fields are unknown but patterns are sought. Examples of tools and processes for each method are provided, highlighting advantages and disadvantages for effective data analysis.
SQL: Enough business intelligence time for administration intelligence
- 1. Enough BI, time for Administration Intelligence! Dr. Nico Jacobs, @SQLWaldorf
- 2. Business Intelligence is not about business • Business Intelligence (BI) is applying tools to business data – Sales data – Stock management – … • But why not apply these tools administration data? – Log files – Status tables – …
- 3. 3 ways to get the Intelligence 1. Reporting – Directly query the data • Fastest to create Query
- 4. When to use reporting? • Reporting is ideal in situations where 1. You know which fields you want to see, and 2. There is only one interesting way to look at it
- 5. Reporting tools Management Studio Reports Ad-hoc queries Activity monitor Database engine tuning advisor … Dedicated tools Idera Redgate Lepide SolarWinds (Confio) … Other tools Perfmon Event viewer Excel Access reports …
- 6. Demo: Custom reports in Management Studio Ingredients: • SQL Server Data Tools Business Intelligence • Management Studio (for SQL specific) Steps • Write a query to read database log file sizes • Run the wizard • Adjust the report to your own taste • Run it in Management Studio
- 7. Custom SSRS Reports • Filtering, grouping and calculating derived values can be done – in the query (if you like SQL), or – in the report (if you like VB.Net) • Easy to add nice visualizations • Integrates with Management Studio • Adopts to context: current SQL Server instance and database • Support print, export and multi-page
- 8. Reporting Disadvantages • Hard to reuse any work: – Every new report requires new, complex queries – Whenever the data source changes, we must update all reports – Certain report constructs can be challenging • Growth figures • Year-to-date calculations
- 9. 3 ways to get the Intelligence 1. Reporting 2. Modeling – Copy relevant data in model – Define all useful calculations once in the model – Reuse them in multiple reports Query
- 10. When to use modeling? • Modeling is ideal in situations where 1. You know which fields you want to see, and 2. There are multiple interesting ways to look at the problem
- 11. Modeling tools • SQL Server Analysis Services Multi-dimensional • SQL Server Analysis Services Tabular • Excel PowerPivot • Pentaho Mondrian • IBM Cognos • Microstrategy • …
- 12. Demo: Building a model in PowerPivot Ingredients • Excel 2010 or 2013 for the modeling and pivot report • Excel 2013 for the PowerView report Steps • Load data into PowerPivot model • Create relationships between tables • Defined calculations where needed • Build reports to taste
- 13. Building models Advantages • Easy to quickly create multiple reports • Allows for advanced analysis • Lots of reuse options Disadvantages • More work before first report is available • Need to master two querying tools
- 14. 3 ways to get the Intelligence 1. Reporting 2. Modeling 3. Mining – Define search space of possible ‘patterns’ – Report on the discovered patterns Query
- 15. When to use data mining? • Data mining is useful when: – You don’t know which fields you want to see! – You do know what type of pattern you’re interested in – You do have relevant historical data
- 16. Small case study: Capacity planning • We could start with very simple model: – 1 variable linear regression, easy 2D chart 0 100 200 300 400 500 600 700 800 900 10 12 14 16 18 20 22 24 26 28 30 Indexes vs mdf size
- 17. Multiple variables • But what if we have more information? – If we know the number of clustered and non- clustered indexes, we must work in 3D… – If we know number of tables as well, we must work in 4D – And if we add columns, blob columns, … we quickly end in 10D or more
- 18. Linear regression • One of the many patterns that data mining can discover is linear regression formulas in multiple dimensions
- 19. Regression trees • If we add to our model non-numeric information (server name, database type), we could build different regression formulas for different situations • This is a regression tree
- 20. Many types of patterns • Regression trees is just one of the many types of data mining available
- 21. Mining tools Microsoft Analysis Services Multi- Dimensional Excel front-end Commercial tools SPSS SAS … Open Source tools R Pentaho Weka Knime …
- 22. Demo: Analyzing Report server log files with data mining Ingredients • SQL Server Analysis Services Multi-Dimensional • Excel 2007 or later with Data Mining add-in Steps • Load relevant data in Excel worksheet • Use Excel Analysis ribbon to select relevant type of pattern
- 23. Conclusions • The Business Intelligence tools are not just for business analysts! • If you know which are the important fields to report upon – And you can only imagine one way to look at it: use reporting – But you can come up with multiple views on these fields: use modeling • If you don’t know which are the important fields in a certain problem, consider data mining
- 24. Get the material • You can download this presentation, report definitions and Excel workbooks from http://1drv.ms/1quLAnw
- 26. Follow Technet Belgium @technetbelux Subscribe to the TechNet newsletter aka.ms/benews Be the first to know
- 27. Belgium’s biggest IT PRO Conference