Moving Forward with the On Demand Real-time Enterprise

Chapter 3. Architectural considerations 91

integration and data movement facilities. But there are still many custom build

integration and data movement solutions in production around the world. These

implementations do play an important role in the organizations that use them, but

as time has passed it has become more and more costly to further develop and

maintain the solutions. Organizations today want to have standard solutions that

can, in a cost effective way, be integrated and isolated so it is possible to easily

replace individual components or entire architectures.

The trend, and the feasibility, of many applications to move towards an SOA

enabled environment has applied pressure on these proprietary custom built

solutions. The market now wants solutions built on standard software, standard

packages, or even built on open source standards which are becoming more and

more a business demand.

3.2.6 Combined approach

Companies seldom choose just one architecture or solution to solve their entire

integration infrastructure. Because of the often different demands for

transformation, volume, velocity, variety and access demands to data, a single

architecture will rarely be able to solve all of the integration and data movement

needs of an enterprise. So it is quite typical to see an enterprise using ETL tools

to solve their ODS and data warehouse data population requirement, and also

using either EAI or EII for ad-hoc information and application integration needs.

ETL versus EII versus EAI versus data replication versus

custom build

We have chosen to list a set of capabilities we find important to consider when

comparing an architecture for a data warehousing environment and a real-time

enterprise requirement. The list is not prioritized by importance and does not

necessarily represent a full list for qualifying which architecture to choose for

which tasks. The comparison of the architectures are described for how the

current technologies are most typically used and not with the regards to

individual product capabilities.

Data Flow

To resolve a data flow all architectures can be bidirectional, even though

traditional ETL has been deployed as one-way data flows. Data replications are

by default designed to be bidirectional, but are often also implemented as

unidirectional.

Data Movement

ETL offers a scheduled batch oriented approach which compares with both EII

running in right-time direct access mode through open access protocols, and EAI