ActiveWarehouse/ETL - BI & DW for Ruby/Rails

NB: This presentation was delivered at the Singapore Ruby Brigade meetup 6-Jan-2010 (at hackerspace.sg)

BI & DW for Ruby/Rails

“!???”

Why should we care about this enterprisey stuff?

Have you heard a client ask for..– A “dashboard”?– Management reports?– Operational statistics?

..in addition to the actual site?

Or maybe you want to pitch for the dashboard/BI projects

themselves?

..using your rails skills of course

BIBusiness Intelligence

CPMCorporate Performance Mgmt

BPMBusiness Performance Mgmt

B&PBudgeting and Planning

EPMEnterprise Performance Mgmt

DashboardEnterprise Dashboards

BI Basics

No, BI is not (always) an oxymoron

Infrastructure & Systems

BI = Business Feedback & Control Systems

Keeping the doors open

Uptime on the

servers; alerts

Infrastructure & Systems

OperationalManagement

BI = Business Feedback & Control Systems

Keeping the doors open

Optimising in the short termintra-day

Focus on systems in isolation

Need extra call centre

staff on shift?

Daily sales

numbers?

Infrastructure & Systems

OperationalManagement

ExecutiveManagement

BI = Business Feedback & Control Systems

Keeping the doors open

Optimising in the short termintra-day

Focus on systems in isolation

Strategic performancemonthly, quarterly, yearly

Across all systems

Profitability by

product

Utilisation and sales

performance

Infrastructure & Systems

OperationalManagement

ExecutiveManagement

Traditional Rails perspective..

e.g. NewRelic

Custom AR reports

Someone else’s problem (opportunity)

Someone Else’s Problem..

Your Rails Storefront App

Fulfillment(maybe a third party)

To report on sales

fulfillment..

AR/AP/GL

To report on revenue and profitability..

To report on sales revenue,

actuals and forecast..

And don’t forgetall those other systems..

CRM MRP

FA

Who is “Someone Else”?

The gigaohm network: “5 Free Business Intelligence Crunchers for Your 2010 Arsenal”

ODS

ETLYour Rails App

Other TransactionalSystems

Data Sources

DBoR, relational reporting

BI & DW

A copy of transaction data specifically structured for

query and analysis

Extract – Transform – Load

Or, Extract – Load – Transform

Or, Transform – Extract – Load

(depending on the technology)

“cubes”

Sales = $22 Customer ID Product ID Date ID …

Customer dimension Date dimension

Product dimension

Fact categorisation“Fact”

MOLAP, ROLAP, HOLAP

MOLAP: proprietary format to optimize for analytical queries

ROLAP: use relational database to mimic multi-dimensionality

HOLAP: hybrid. Drive analytics from MOLAP, drill down to relational

Star schema Snowflake

Why?? What’s wrong with..

select a.name, sum(b.amount) from products ajoin order_items b on a.id =b.product_id group byproduct_id

Product.sum(:amount, :include=> :orders, :group =>‘product_id’)

•Every question needs it’s own query

•Can’t predict all the questions in advance

•Un-scalable grunt work

ActiveWarehouse ActiveWarehouse-ETL

ActiveWarehouse

Rails plugin by Anthony EdenROLAP solution based on ActiveRecordFeatures

– Generators for Facts, Dimensions, Cubes and Bridges

– Supports calculated fields– View helpers for reports with drill down

ActiveWarehouse-ETL

Rails gem/plugin by Anthony EdenDSL for extract – transform – loadSource/sink: file, db, xml, .. (extensible)Features

– Pre/post processors– Transformations

The Cupcakes Store

Use Activewarehouse-etl to load seed data from csv to app db (mysql)

1

The Cupcakes BI Dashboard

2Use Activewarehouse-etl to load dimension and fact data to the warehouse(mysql to mysql)

3Use Activewarehouse to build a simple analytical “dashboard” and reporting tool

Follow the documentation at http://github.com/tardate/cupcakesinc to see how this works (and try it yourself)

Product listing at Cupcakes Inc..

Customer listing at Cupcakes Inc..

Order listing at Cupcakes Inc..

Order detail at Cupcakes Inc..

Sales By Product AW Report

Sales By Product (drill to 2009)

Reasons to be Cheerful..

Language

ETL processing, cube rules etc typically use custom languages (often archaic and limited)

BISuites

It’s … ruby!

UI Customisation and Presentation Integration

Web delivery typically very constrained.

Often rely on strong integration with office software (Excel). Leads to “custom application development in Excel” syndrome.

BISuites

It’s … ActionPack!

Google maps mashups, social graph links. .. you get full UI control, as long as you have the development budget.

Speed of development

Basic deployments can be very fast.

But UI inflexibility can lead to either lots of time wasted trying to shoe-horn, or need to “reset customer expectations”

BISuites

It’s … Ruby & Rails. Say no more ;-)

TCO

Top-tier suites can come with a hefty $ tag. And prices are going up..

But some analysts are predicting 2010 to be the year BI gets FLOSS momentum (see gigaohm review of 5 well established alternatives)

BISuites

It’s … Ruby & Rails. Say no more ;-)

Trade-in software license costs for more development.

Caveats..

Native MOLAP

Generally good support for database MOLAP features.

Can be platform specific though – e.g. Microsoft MDX, SQL Server Analytical Services

BISuites

A gap. No real support currently available.

ActiveWarehouse uses relational model to “fake” MOLAP (ROLAP)

Performance

Generally, all established analytical engines (and backing databases) have great performance track record. Huge scalability (millions of rows)

BISuites

Unproven. ActiveWarehouse/ETL does not have many (public) proof points.

Given that it is tied to AR performance, expect scalability could be an issue.

Take-aways~ActiveWarehouse

It’s an impressive codebase. When you get it working, it works well.. but

– Virtually no documentation!– No contemporary examples– Not under very active development– A “textbook” data warehouse implementation.

May or may not be exactly what you want..

Remember: – data is batched. Not realtime.– Rails 2.x : install the plugin (gem is 1.x)

3

Take-aways~ ActiveWarehouse-ETL

Neat tool. In addition to feeding AW:– Generate and load seed/test data– Move data between systems

But again,– Poor documentation– When it fails, can do so silently (makes

sure filename paths are delimited correctly for your platform!)

2

Take-aways~ BI on Rails Solutions

Plain AR– just avoid the rabbit hole

AR + ETL– get all the data you need in one place

AW+ETL– traditional ROLAP, make Rails the focus of the BI effort

Go the BI suite route– When you need to adapt to many transactional systems

at scale, and customer has the $$ – (Rails remains just for transactional apps)

Or… (discussion point;-)

1

Thank you!

Questions?

0

Some References

ActiveWarehouse: http://github.com/aeden/activewarehouse ActiveWarehouse-ETL: http://github.com/aeden/activewarehouse-etl Cupcakes Inc sample site(s): http://github.com/tardate/cupcakesinc Singapore Ruby Brigade (SRB): http://groups.google.com/group/singapore-rb