By using website you agree to our use of cookies as described in our cookie policy. Learn More


Expanding Business Intelligence Across the Enterprise

By Robert D. Schneider

Until fairly recently, a comparatively small number of users have been able to realize the full potential of business intelligence and data warehousing (BI/DW). These users have typically worked with relatively isolated data sets, with minimal sharing and interaction across departments. However, a new class of enterprise-scale BI/DW solutions is bringing this isolation to an end.

Why Business Intelligence and Data Warehousing Need More Horsepower

Today’s BI/DW solutions are the culmination of decades of advancement. The relatively primitive operational reporting products of the 1970s and 1980s eventually matured into the first business intelligence solutions in the 1990s and early 2000s. Modern business analytics represents the next step of this evolution, but a collection of arduous impediments are combining to delay or derail this momentum:

  • Usage expansion
  • Big data
  • Data diversity
  • Complex questions
  • Decision velocity
  • IT resource rationing

The only way to surmount these obstacles is to embrace the reality that it is no longer acceptable for your BI/DW solution to restrict the amount of data, types of workloads, and number of concurrent users. To fully exploit the power of business analytics, you must do two things:

  1. Ensure that all your enterprise’s data and analytic algorithms are reachable through a single, far-reaching BI/DW information grid.
  2. Select and implement a flexible BI/DW solution that is capable of working with many concurrent users and all possible applications.
Guidelines for Selecting a Modern BI/DW Solution

Although each enterprise is unique, there are certain facts and best practices that should apply in just about every situation:

The BI/DW software must provide a central source of information that addresses all analytics and reporting needs. There’s no way to get the scale you need from a collection of specialized, fragmented BI/DW systems. Instead, all relevant resources should be grouped, managed, and presented to users as unified assets, with support for a broad range of analytic and reporting applications. This makes data and processing power available throughout the entire enterprise.

The BI/DW software must handle enormous data volumes and variety. Relational databases will continue to store a significant percentage of your information and should support both structured and unstructured data. Your goal should be to deploy a single solution capable of working with all your enterprise’s data. Also, because storage is cheap but not free, your selection must be able to leverage the power of natural compression. Finally, to help diminish latency, all data and related analytic algorithms should be kept close together.

The solution should deliver results quickly. Because information becomes stale so fast, it’s incumbent upon you to provide answers to your users’ questions as quickly as possible. Well-proven patterns such as massively parallel processing (MPP) make accelerated response times possible. For example, MPP lets you scale out across multiple heterogeneous commodity servers, which enables effective workload balancing. Understand the different approaches in the marketplace: shared-nothing MPP and shared-everything MPP. Both scale out for querying massive volumes of data. Shared-everything architectures are good at scaling to support many concurrent users, making optimal use of storage resources, and simplifying administration.

Costs should be kept to a minimum. Because budgets aren’t keeping up with the demands being placed on IT, your only option is to squeeze the most you can out of your existing resources. For example, scale-out architecture can leverage commodity servers, which lets you decommission costly, proprietary platforms and eliminate duplication of data and unnecessary maintenance of many data marts. You can also employ a collection of complementary strategies to reduce administrative overhead:

  • Virtualization
  • Automatic workload balancing
  • Self-tuning systems
  • Well-integrated management tools

The days of isolated BI/DW solutions servicing a mere handful of concurrent users are drawing to a close. Information and analytic algorithms will be incorporated into central grids of compute and storage assets that can scale to support the entire enterprise. If you understand your needs and apply the right set of guidelines and best practices when selecting an enterprise-grade BI/DW package, you’ll bring the power of analytics to a much broader audience within your enterprise.

This article originally appeared in the issue of .

TDWI Membership

Get immediate access to training discounts, video library, BI Teams, Skills, Budget Report, and more

Individual, Student, and Team memberships available.