life in hell: the experience of successful bi managers

23
1 What's Required to Grow an Enterprise BI Deployment? December 7, 2010 Mark Madsen www.ThirdNature.net

Upload: mark-madsen

Post on 27-Jan-2015

105 views

Category:

Technology


1 download

DESCRIPTION

Life in Hell: The Experience of Successful BI Managers - retitled from the boring original. Growing business intelligence from a departmental deployment or pilot project to an enterprise scale is not as simple as buying more licenses. Increasing the scope of BI means taking into account a broader set of information needs in the organization, increasing the size and variety of data needed in a warehouse or mart. Broader use implies more varied delivery needs, as well as scaling up the user count. Increasing the scope of use also raises the importance of managing day to day operations and reliability. This webcast will discuss factors to consider when increasing the scope of a BI program, and what requirements that puts on components of the technology infrastructure. Different axes of enterprise scale How information delivery requirements change as you expand BI The impacts of growth on operations and administration Slides from a webcast for TDWI. You can listen to the full webcast and see the Jaspersoft/Talend presentations here: http://ow.ly/3lkHj

TRANSCRIPT

Page 1: Life in Hell: The Experience of Successful BI Managers

1

What's Required to Grow an Enterprise BI Deployment?

December 7, 2010Mark Madsenwww.ThirdNature.net

Page 2: Life in Hell: The Experience of Successful BI Managers

Maturation of BI

A healthy BI program gets more complex and harder to manage over time.

Growth from the initial installation to broader organizational use has many aspects:▪ Number of users

▪ Multiplying uses

▪ More information

Growth requires adapting processes and technology.

Page 3: Life in Hell: The Experience of Successful BI Managers

Maturation: Initial Build, the Early Stage

Projects

ROI Minimum ROI hurdle for initial project to be built, infrastructure created

Future work planned for

Done

Consultants leave here

Copyright Third Nature, Inc.

Page 4: Life in Hell: The Experience of Successful BI Managers

The Origin of BI Backlog: Next Phase

Projects

ROI

Minimum ROI hurdle is lower for subsequent work

Done

Fewer resources, so work takes slightly longer to complete, but not so long as initial build

Copyright Third Nature, Inc.

Page 5: Life in Hell: The Experience of Successful BI Managers

The Long Tail of BI

Projects

ROI

This is what happens to successful data warehouses

Done

“Oh crap”

Copyright Third Nature, Inc.

Page 6: Life in Hell: The Experience of Successful BI Managers

Prioritizing the Long Tail of BI

Projects

ROI

Financial priorities, business priorities, steering committees, budget limits, time-boxing…(Guess which things get done)

Copyright Third Nature, Inc.

Page 7: Life in Hell: The Experience of Successful BI Managers

Prioritizing the Long Tail of BI

Projects

ROI

Executive pet projects

Financial priorities, business priorities, steering committees, budget limits, time-boxing…(Guess which things get done)

Copyright Third Nature, Inc.

Page 8: Life in Hell: The Experience of Successful BI Managers

The Long Tail of BI: Why We Have Spreadmarts

Projects

ROI

The Kingdom of ExcelLow hanging fruit

*sigh*

Mismanage this process and you have a legacy system everyone complains about

Copyright Third Nature, Inc.

Page 9: Life in Hell: The Experience of Successful BI Managers

Requirements, data sources

ETL / DI

Warehouse / Mart

BI / Analytics server

Clients

Development Process Designed to Minimize Later Change

Metadata stored here, almost but not quite the same each time

The common MD repo was supposed to fix this

This process is fine for the initial build.

Three months later, not so much.

Page 10: Life in Hell: The Experience of Successful BI Managers

Two Things People Don’t Want

Data integration and BI projects that take months to deliver for business needs that may be one‐time or done in weeks.

Least‐common denominator financial and transaction data with contextual information and details stripped away in the name of speed.

Slide 10

Page 11: Life in Hell: The Experience of Successful BI Managers

The Process is Not a Waterfall, It’s an Ongoing Cycle

The BI layer is the starting point for users, not the end point.

As people adopt new information, needs alter, driving change.

Your processes switch from “build” to “keep it running”.

Requirements, data sources

ETL / DI

Warehouse / Mart

BI / Analytics server

Clients

Slide 11Copyright Third Nature, Inc.

Page 12: Life in Hell: The Experience of Successful BI Managers

Growth: Increased UseIf you’re successful, users become more proficient and BI use increases.

Effects are:▪ Performance problems

▪ Capacity problems

▪ Shrinking data load windows and response time requirements

This presages more user growth.

Page 13: Life in Hell: The Experience of Successful BI Managers

Raises problems of scaling not only performance, but also managing accounts, data security, licenses.

Aspects of Growth: More Users

Page 14: Life in Hell: The Experience of Successful BI Managers

Growth: Increased Variety of Uses

Once people master the basics, diversifying demands require new tools, more complex analysis and models.

Page 15: Life in Hell: The Experience of Successful BI Managers

Warning about software vendors:

The Swiss knew when to stop. Vendors often don’t.

BI and DI vendor response has been to add features to the tools to meet all the different use cases.

Different uses can drive conflicting tool requirements.

Page 16: Life in Hell: The Experience of Successful BI Managers

Growth:  Data

Data is the item everyone focuses on when talking about growth and BI.

The primary impact is on the database, both getting data out and getting data in.

Data volume is the easiest problem to address (in most common BI / DW situations).

Page 17: Life in Hell: The Experience of Successful BI Managers

Other aspects of data growth are harder to address

Variety:1.More sources▪ More system types

▪ APIs and other oddities

2.More types▪ The usual suspects

▪ And more

3.Uses that require more complex data models or transformations.▪ Like data mining

▪ And sandboxes

Page 18: Life in Hell: The Experience of Successful BI Managers

Growth often drives the need for lower latency12x5 moves to 24x7 operation, driving SLAs, capacity planning, failover and disaster recovery.

Methodology, organization and technology are related. Speeding up one won’t always speed up the others.

18

Page 19: Life in Hell: The Experience of Successful BI Managers

New Data Integration Methods and Tools Required

The initially designed ETL may not handle the varied data latencies, or deliver the performance to meet smaller batch windows.

Daily+

Single batch

Frequent batch

Continuous load

Streaming

Immediate

Mini‐batch

Copyright Third Nature, Inc.

Page 20: Life in Hell: The Experience of Successful BI Managers

Development, Maintenance & Operations

Real time decisions on low latency data mean data quality plays a larger role, and it’s harder to address.

Warehouse availability becomes much more important to the business, and it isn’t just the database – it’s everything.

Performance and meeting strict BI SLAs will rise in importance since you are now tied in to business operations.

Slide 20

Page 21: Life in Hell: The Experience of Successful BI Managers

Administration

As BI grows in importance within the organization this becomes a focal point.▪ Any problem is magnified due to the broader scope.

▪ There are more products.

▪ There are more inter‐tool dependencies and problems are more distributed.

▪ The least‐emphasized set of features in most BI tools.

▪ And most DI tools.

Page 22: Life in Hell: The Experience of Successful BI Managers

Overall Effect of Growth

Complexity, which leads to• Performance problems

• Reliability problems

• Maintenance problems

• Difficulty adapting to change

Usually we forego features useful in the long term during our product evaluations in favor of features important for initial delivery.

Page 23: Life in Hell: The Experience of Successful BI Managers

About the Presenters

Mark Madsen is president of Third Nature, a technology research and consulting firm focused on analytics, business intelligence and data management. Mark is an award‐winning author, architect and CTO whose work has been featured in numerous industry publications. He is an international speaker, a contributing editor at Intelligent Enterprise, and manages the open source channel at the Business Intelligence Network. For more information or to contact Mark, visit  http://ThirdNature.net.