Chapter 3. Application Integration pattern 67
Is the enterprise data topology centralized or decentralized?
– Centralized: This integration effort will bring about centralized access to
all or a subset of the enterprise data model.
– Decentralized: Applications will retain their isolated repositories, but now
will have cohesion based on data integration.
What is the database affinity type?
– Homogeneous: All repositories are of the same type.
– Multi-vender Relational: All repositories are relational with ODBC/JDBC
support for interoperability but are from different vendors.
– Heterogeneous Structured: Repositories are not all relational, but all
have a structured layout.
– Structured/Non-Structured: This addresses the need to integrate
non-structured (for example, free-form text) with structured data sources.
Refer to the IBM Patterns for e-business Web site for further details:
http://www.ibm.com/developerWorks/patterns
3.6 Previous Application Integration patterns
Table 3-3 on page 68 provides an overview of the relationship between the
previous Process-focused Application Integration patterns and the revised
Process-focused Application Integration patterns presented in this chapter:
Direct Connection is retained for application coordinated requests.
Transactional is now a quality of service. Transactionality may apply to all of
these patterns, so it is applied as a quality of service rather than being a
separate pattern.
Aggregator/Broker are combined into Broker for broker-coordinated requests.
Managed Process is split into Serial Process and Parallel Process for
process-managed coordinated requests.
The read-only versus read/write classification used with old patterns is not
used with the new patterns, since:
– For Transactional and Managed Process, read-only is not applicable.
– For Direct Connection, the same pattern applies in both cases.
– For Aggregator/Broker, the observed patterns are identical.