Is Data Abstraction Preferred Above Data Replication, When Considering BI?

In practice what happens is that when you don't have data abstraction, you have no BI! If BI is all about looking at every transaction, there may be no intelligence in it!

The funny thing is that when you have data abstraction and good BI, users wants to drilldown from the higher levels to individual transactions! So you need Data Replication also.

Data Replication is funnily used not only as a backup database but also to separate transactional systems from reporting systems. Many companies prefer running BI systems against a replicated database leaving the main database only for transactional use!

This is where enterprises are beginning to understand the key differences between Operational BI, Strategic or classical BI and Reporting.

Operational is all about Transactional and real-time data – Data Abstraction , Strategic BI works more off data marts and data ware houses – More on the replication – time window is longer and reporting feeds from the two as well as other data sources!

So it’s all about what you want to achieve Operational or Strategic insight!

To stay ahead of competition, enterprises today are rapidly moving towards deploying consistent and up-to-date BI (Active or Near Real-time BI) solutions throughout the enterprise infrastructure.

Strategic/Traditional BI is mostly focused towards analysts to make informed decisions while Operational BI concentrates towards front-line representatives to make day-to-day decisions. The resulting combination of these two can be rather powerful:

Data Abstraction enables us to exhibit only the bare minimum number of characteristics/features of the subject area under consideration i.e. a glimpse of "the big picture". This in turn requires data integration, and application of certain business rules to obtain the final abstraction. All these can be comprised under Strategic BI. Parameters such as "Data freshness" and "Change Data Capture" can be important deciding factors here. While Data Replication, on the other hand, would provide access to most granular facts and is the next step once an analyst has looked at the abstracted features to validate the predictions against historical trends, say curve fitting on newly arrived granular data. A just-in-time approach.

Systems of the types SOA, web services, work-flow and supply chain demand replication. This would require more storage/memory, real-time integration, etc. The DW must support a mixed workload of real-time data loads, complex reporting, and even tactical requests simultaneously. What enterprises would need is the ability to get an "integrated" "single view of the truth" "at the right-time" "for all users". So the best choice would be a combination approach (Pervasive BI) depending upon business needs, where it can be separated by a logical abstract data model.