Advisors:

The amount of data being processed nowadays is continuously increasing. This fact also applies to data in databases. Modern computer systems have a large variety of het-erogeneous processing capabilities, such as (multiple) CPU(s) and (multiple) GPU(s) or other co-processors. Often, the database management systems do not use the extra processing capabilities of co-processors and completely rely on processing on the CPU. But several authors have already shown that using the additional processing capabil-ities from co-processors can significantly speed up the query processing on database management systems under certain circumstances. One of the factors influencing the query processing on heterogeneous processor environments is the operator placement. This means, it is important to decide which operation of the query is executed on which processor. Current approaches use greedy heuristics or specialized approaches which aim to optimize for a certain goal, e.g. execute complete paths on a certain co-processor. Classical optimization approaches, such as Simulated Annealing or Genetic algorithms have not been evaluated so far. Therefore, this thesis concentrates on evaluating oper