Optimierung datenintensiver Workflows: Konzepte und Realisierung eines heuristischen, regelbasierten Optimierers

Nowadays, workflowand database technology is crucial for automating business processes. Workflow management systems execute business processes by means of workflows that integrate numerous application systems and data sources. Typically, database management systems store business data that is referenced by business processes. Obviously, the performance of a business process is directly dependent on the efficiency of data access, data processing, and data management. Leading vendors of workflowand database management systems extended workflow description languages like BPEL by SQL-functionality. This simplifies the modeling of dataintensive workflows that process huge amount of relational data. Thereby, data processing operations like SQL statements or stored procedure calls may be defined directly on the workflow level, instead of encapsulating them into Web services. Thus, a workflow description discloses the whole optimization potential concerning its data management. The outcome of this is a new query optimization approach that completes existing approaches in database systems: Rewrite rules transform inefficient data processing operations in a workflow description into equivalent operations that perform better on a workflowand database management system respectively. This results in an improved performance of data-intensive workflows with respect to their data management. This doctoral thesis introduces a heuristic, rule-based optimizer for data-intensive workflows. In order to improve the data management of a data-intensive workflow, the optimizer applies an appropriate rule set according to a well-defined control strategy on an internal representation for data-intensive workflows, the so-called process graph modell (PGM). PGM is adjusted to the optimization of data-intensive workflows and is the appropriate basis for rule-based transformations. It allows for an exact and language independent definition of rewrite rules. Consequently, it supports various optimization scenarios covering a multitude of data processing operations defined in different description languages. The rule set contains rewrite rules that are based on existing as well as on new optimization techniques. Each rewrite rule consists of two parts: a condition and an action part. The condition part defines what conditions have to hold for a rule application in order to preserve the original workflow semantics. It refers to control flow dependencies as well as to data and communication dependencies. Additionally, it considers detailed information of activities. The action part of a rewrite rule defines the transformations applied to a workflow description provided that the corresponding condition part is fulfilled.