Cost-Based Optimization of Integration Flows - Datenbanken ...

2.1 Integration Flows
Executing Integration Flows
Semantics
Data Flow
Control Flow
Data
Granularity
iterator
model
hybrid
instanceglobal
materialized
intermediates
materialized
intermediates
Figure 2.6: Classification of Execution Approaches for Integration Flows
instancelocal
instanceglobal
instancelocal
instanceglobal
instancelocal
instanceglobal
Putting it all together, Figure 2.6 illustrates the overall classification. Using controlflow-oriented
execution semantics (with temporal dependencies) directly implies the use
of materialized intermediates in the form of variables. In this context, both instance-local
and instance-global processing is possible. Typically, EAI servers and BPEL engines use
this execution approach. Hence, this thesis uses the execution model of control-flow semantics
and materialized intermediates as conceptual foundation. Furthermore, data flow
execution semantics allow for both the iterator model and materialized intermediates as
well as both instance-local and instance-global data granularity. Examples for the use
of materialized intermediates are Demaq [BMK08] (instance-global) and some ETL tools
(instance-local). The iterator model requires a more fine-grained classification. Iterator,
instance-global refers to a tuple-based processing over data of multiple instances, where
punctuations are used to distinguish data from the different instances. An example for this
model is the stream-based Web service approach [PVHL09a, PVHL09b]. In contrast, iterator,
instance-local refers to a tuple-based processing over data of a single instance, which is
the typical execution model of ETL tools. In addition, an iterator, hybrid instance-global
model can be used, where the single materialized intermediates of multiple instances are
executed in a pipelined fashion and thus, with the iterator model.
Finally, we will use this classification of execution approaches in order to position the
different optimization approaches as well as the results of this thesis.
2.1.5 Optimizing Integration Flows
Based on the different modeling and execution approaches, we now focus on the optimization
of deployed integration flows. The main scope is the optimization of integration flows
with control-flow execution semantics.
Due to the early state of the research area of integration flow optimization, an exhaustive
classification of optimization approaches for integration flows does not exist. However,
typically, integration-flow-specific characteristics are exploited for optimization. First,
the expensive access of external systems is tackled with approaches that speed up the
external data transfer. Second, the control-flow-oriented execution is optimized by parallel
execution of subplans of an integration flow or by operator reordering. Thus, we use these
two categories in order to classify the existing approaches.
15