Scheduling in Grid Databases

Many global organizations are generating large volumes of highly distributed data that must be shared for querying by the organization´s participants, which can be put together by a grid infrastructure. Scheduling query execution in these environment is an important issue. Centralized and hierarchical scheduling architectures are between the most commonly used scheduling architectures in distributed databases and grid computing, respectively. Each architectural model has its own advantages and disadvantages. In this paper, we evaluate the usage of both models in querying highly distributed data. We consider several scheduling policies and experimentally compare scheduling models and policies in terms of throughput and SLO-fulfillment. The obtained results indicate that layered on-demand scheduling may lead to good results in both throughput and SLO-oriented environments.