Refresh Technology On Workflow Of Database Using Provenance

With increasingly sophisticated operations on the database, the workflow technology has been introduced in a series of operations on data, which can execute data operations automatically according to the operating procedures designed in advance, so that the database-oriented workflows appear. As to database-oriented workflows, the refreshment of data is one of the key issues which we must to face with, the changes of input data will be bound to affect the other data on the workflow. When the workflow has large amount of data and many data manipulations, the traditional refresh technology is hardly to provide a satisfactory refresh rate, so we need a real-time refresh technology which can update the data on the workflow efficiently. For this, the technology of provenance management is introduced into the workflow, and someone has put forward a kind of selective refresh method based on provenance to update the specified output data tuple. This method back-tracks from the output data which needs to be refreshed according to its provenance information to find the relevant input data, and rerun this part of the input data to refresh the data of the workflow.Based on the existing work, two refresh methods of workflow are proposed in this paper: directly selective refresh method and step-by-step selective refresh method. The directly selective refresh method is to record the relevant input data tuples using provenance information, and when the ouput data tuples need to refresh, it can find out the relevant input data directly according to its provenance information, so that the time of tracking operation will be shorten and the speed of refreshing the workflow will be improved. The step-by-step selective refresh method is to update the data forword from the updated input data according to the operations of the workflow and the provenance information stored in data tuples, and gradually selectively refresh the entire workflow. That method takes advantage of the intermediate data which has been stored, and it further reduces the quantity of rerun data during refreshment and improves the efficiency of refreshment, what’s more, it solves the issue that the relationship between data is affected by updated data, and the issue that the data can not been updated when new data tuples were added into the input table. Finally, we conducted a large number of compararison experiments on data sets using different refresh methods. The analysis of the experimental results prove that the two refresh workflow methods proposed in this paper have a higher refreshing efficiency, and the correctness of the updated results on safe workflow with the two methods and on unsafe workflow with the step-by-step selective refresh method.