Dynamic Modification of Continuous Queries

Nowadays, sophisticated near-real time data stream queries run on massive and continuous amount of data, resulting from transactions and measurements. However, these Long-Running Data Stream Queries might not correspond to user requirements at some point and need to be modified. Considering the distributed, continuous and near-real time nature of data streams, modifying a query needs to be conducted through a reliable model. So far, there has not been any model developed by the streaming community to classify continuous query modification methodic. Hence, in this thesis we will introduce a new model to capture the problems rising when modifying a long-running query on a data stream whereas fulfilling the application’s correctness requirements. This model is based on a simple start-stop model, and gets more complicated resulting in several change models. Each of these change models tries to cover some correctness criteria. There are basically two groups of correctness criteria: Safety and Liveness. The former minimizes loss, duplicity and disorder among output elements and the latter guarantees progress of the new version of the query and termination of the old one. I would like to dedicate this thesis to Michel, who made it possible for me to come to Zurich and study at ETH, and to my dear mother for her love and continuous faith in me.