Optimization of execution plans in information mediators is a critical task, specially when sources are remote and semistructured, as in the case of transactional web sites. Lack of Quality of Service in HTTP connections does not ease the measurement of the mediated query performance. Different solutions based on relational and hierarchical optimization techniques have been proposed, mainly in terms of time and space. This article integrates some of these techniques by using a specialized cost repository but, more importantly, defines a physical layer optimization cost model by means of a configurable pool of browsers , which improves dramatically query execution time in web mediators.