Tech2Tech Table of contents Ask the expert Add capabilities to the data warehouse dynamically with Teradata's extensibility features. An active user community Empowering users of business intelligence adds ROI for your enterprise and confidence for your team. Enterprise BPM Integration of technologies from Hyperion and Teradata result in an enterprise architecture for Business Performance Management. Purpose-built platform Enterprise data warehousing requirements drive the design and architecture decisions of the Teradata platform. Tech support Need technical help? A Teradata Certified Master answers your queries.

Print	Send to colleague
PDF (123 kb)	E-mail us

Enterprise BPM

Hyperion and Teradata collaborate on next-generation application architecture.

by Stephen Brobst and John Kopcke

Business intelligence (BI) is integral to a new generation of business performance management (BPM) application architecture. A key characteristic of this application architecture is to provide an enterprise platform for BPM rather than a departmental solution. To achieve this goal, new levels of scalability are required at the point of integration between the BI platform and the underlying data delivery platform. Hyperion and Teradata have jointly undertaken creation of such an architecture.

Through integration of best-of-breed technologies and developing an enterprise framework for data delivery, Hyperion and Teradata have broken new ground in the area of Enterprise BPM. This article describes the approach that we have taken and the resultant solution architecture.

Foundational principles
There are five driving principles that have guided the joint work between Hyperion and Teradata:
1. Extreme ease of use
2. Single source of truth for enterprise decision making
3. Extreme performance
4. Ability to access the detailed data
5. Complementary leverage of technology platforms

These principles provide the foundation for implementation of the joint Hyperion/Teradata architecture for Enterprise BPM.

Extreme ease of use in an enterprise architecture provides knowledge workers in different parts of an organization with the right set of tools for accessing information. The right tools for the right individuals with access to the right information within the enterprise.

Highly technical developers require tools that provide a high-powered SQL access mechanism with a programming interface whereas a business analyst will want an easy-to-use Excel interface with drag-and-drop query and analysis capability. Business users will want dashboards and interactive reports with drill-down capability to the detailed data.

The data warehouse provides a single source of truth with integrated operational, customer and financial data for Enterprise BPM. Scorecarding, budgeting, planning and enterprise consolidation applications with trusted solutions are all integrated on top of the enterprise single source of information for decision making embodied in the data warehouse. Auditable drill paths are provided from dashboards all the way down to the lowest level of detail stored in the relational data warehouse repository.

Extreme performance means that access to information is immediate. Batch reporting with overnight delivery provides an ability to manage the status quo, but it does not allow truly innovative approaches to BPM. Interactive access to data provides an effective means of performing what-if analysis and allowing a proactive (rather than reactive) approach toward BPM. The iterative decision making facilitated by immediate access to information is well proven to provide a superior capability for quantitative analytics.

Previously, delivering extreme performance typically meant using summary data rather than providing access to the detailed data. The problem with this approach is that the detail is usually where the breakthrough answers are to be found. The solution architected by Hyperion and Teradata is designed to provide extreme performance and access to the detailed data.

Figure 1 (click to enlarge)

The Hyperion/Teradata Enterprise BPM architecture is the integration of best-of-breed technologies and an enterprise framework for data delivery.

Through use of optimized hybrid online analytic processing (HOLAP) in the Hyperion application architecture, along with advanced indexing structures and the scalability of the Teradata Database, a new level of performance delivery is possible for BPM. The integrated solution removes all practical limits related to the level of detail that can be used in enterprise analytics with Hyperion.

Complementary leverage of the Hyperion and Teradata technology platforms in an integrated solution provides the basis for the performance, usability and speed to delivery of the jointly developed BPM solution. Through optimization of SQL generation from Hyperion for the Teradata platform, along with advanced index management for Teradata provided via Analytic Integration Services, unmatched performance is delivered for BPM.

The integration of the Hyperion application framework with Teradata's normalized database structures is achieved using a semantic data model layer to deliver easy-to-use access on top of a flexible set of underlying table objects. The use of Teradata's industry logical data models (especially the financial management and customer/party subject areas) provides a prebuilt starter kit for deployment that allows very fast delivery in today's increasingly competitive business environment.

Online analytical processing
It is critical to understand that online analytical processing (OLAP) is a characterization of (analytical) applications, not a database implementation technique. The so-called FASMI test (fast, analytical, shared, multi-dimensional and informational) published by Nigel Pendse and Richard Creeth in The OLAP Report (www.olapreport.com/fasmi.htm) over ten years ago is often used to describe the OLAP framework.

In the past, OLAP was often seen as synonymous with cube implementations. However, advances in technology provided by Hyperion and Teradata allow effective deployment of BPM with relational OLAP (ROLAP) technology.

The ROLAP approach removes the scalability limitations of the multi-dimensional database structures implemented as cubes in the multi-dimensional online processing (MOLAP) approach that prevents providing access to detailed data. The integration between Hyperion and Teradata overcomes the performance barriers to OLAP often associated with relational implementations. Moreover, the use of the semantic layer in the Hyperion/Teradata solution provides a dimensional model without sacrificing the flexibility of a normalized database implementation.

Hyperion's implementation of HOLAP technology allows the power of the BPM application to access the enterprise data warehouse directly in Teradata transparent to the end user. The semantic model provided is one of a cube, but the Hyperion application will translate the end-user information request into SQL for accessing the Teradata Database. This is completely transparent to the end user.

Hyperion integration with Teradata allows information access to either Essbase cubes or the Teradata relational database—with no difference in user interface for the knowledge worker.

This implementation of HOLAP combines the best of both worlds. For less volatile data, pre-aggregation using Teradata can be used to feed into Essbase cubes to allow very efficient access to summarized data.

For more quickly changing data, or when more detail is required, Hyperion provides direct access to the Teradata Database. The Hyperion application will access the appropriate storage repository (Essbase or Teradata) automatically.

The DBA chooses the appropriate storage repository based on data volumes, access characteristics, volatility and overall business requirements to determine what goes into the Teradata relational repository and what goes into the Essbase cube. The DBA then configures the Analytic Integration Services suite's OLAP meta-outline for hybrid/relational access and the Hyperion application engine takes over from there. The choice of storage repository is completely invisible to knowledge workers; end user look-and-feel is identical regardless.

Data model deployment
Practitioners in the data warehousing industry frequently confuse construction of the semantic data model, logical data model and physical data model. Each of these constructs is critically important in the successful deployment of an enterprise data warehouse solution. And yet they serve very different purposes in the implementation of an enterprise solution.

A semantic data model (SDM) captures the business view of information for a specific knowledge worker community or analytic application. A logical data model (LDM) captures the business relationships in the enterprise information independent of a specific analytic application or departmental view. A physical data model (PDM) captures the implementation design of tables in the data warehouse.

There can be many semantic data models for the various knowledge worker communities in the enterprise, but there is only one logical data model and only one physical data model in an enterprise data warehouse implementation.

The logical data models provided by Teradata in each industry use a third normal form design philosophy for deployment of an entity-relationship model. Engineering best practices are then used for selective denormalization, index selection, partitioning and other translations from the logical model to the enterprise physical data model for the warehouse.

Semantic data modeling involves understanding the business requirements for a specific knowledge worker community. The Hyperion BI platform maps into the Teradata PDMs using a semantic data model as the bridge between the business view of the information and the underlying technical implementation of tables in the relational database.

Semantic data models are often implemented using dimensional modeling techniques. Dimensional modeling is a more restrictive form of entity-relationship modeling wherein many-to-many relationships are not allowed in the end user's view of information. All relationships are mandatory many-to-one and only a single path is allowed between any two levels in a dimensional hierarchy.

These modeling rules ensure enforcement of the MECE principle. MECE means that all metrics are presented as mutually exclusive and collectively exhaustive in the analytic framework. No dollar will be lost and no dollar will be double-counted.

Once designed, the dimensional SDM is deployed on top of the Teradata PDM using views. Business rules are used to disambiguate any many-to-many relationships in the semantic data model implementation. In this way, the underlying physical data model retains the power and flexibility of many-to-many relationships while still allowing ease-of-use for analysis via the dimensional model captured in the SDM layer. The Analytic Integration Services tool maps directly into the semantic data model implemented on the Teradata data warehouse.

The Analytic Integration Services tool can also be used for creating the SQL to build aggregate join indexes (AJIs) on Teradata. Selective use of AJIs can be a valuable tool for extending the underlying physical data model to deliver extreme performance when accessing key performance indicators as aggregate values using relational OLAP techniques in Hyperion's BPM application.

AJIs are also used to accelerate cube build times for Essbase when an advanced relational access strategy is undertaken. Aggregate awareness allows Hyperion to generate SQL that will undergo automatic transformation via the query re-write engine within the Teradata Database. Moreover, AJIs are automatically maintained by the database upon record insert, update or delete operations—so manual maintenance tasks required from DBAs associated with traditional summary tables are eliminated.

Conclusions
BPM provides the context to make BI meaningful within large organizations. Integration of best-of-breed technologies from Hyperion and Teradata provides an unbeatable capability for high-end analytics in this space. Delivery of this solution has been a multi-year effort on the part of the engineering organizations in both companies and is the result of a shared vision for enterprise application architecture for advanced business intelligence. T

Stephen Brobst is the chief technology officer for Teradata.

John Kopcke is the chief technology officer for Hyperion.

© Teradata Magazine-March 2006

RELATED LINKS:

Teradata and Hyperion: Transforming Enterprise Data Into Business Performance