Workflow Implementation On Grid Data Mining Platform

This article is about a workflow design which is based on the data mining platform (BillionGrid). Each user request for data processing is handled by a call to a data mining algorithm in BillionGrid platform. This approach for the analysis of complex data is too simple, which can't meet people's requirements of processing the complex data.In order to do a more in-depth and more flexible analysis on data, in this paper, shows a way which is build a workflow on the grid data mining platform to handle data. Workflow allows users to handle data in process, which greatly increased the flexibility and the complexity of data handling. Meanwhile, it must be easy to use, for this reason, this workflow is designed in three-tier. The users use a graphical interface designing workflow in the first layer, which reduce the difficulty of the designing. In this layer the workflow is designed by BPMN, and the workflow design ideas come from Intalio Designer, this layer is implement by bxmodeller.The concrete implementation of the workflow is in the second layer, this layer processing workflow tasks by system which is transparent to users. The main work in this layer is to translate the BPMN in the first layer to BPEL workflow language, and then release BPEL workflow to web service, then use wsdl2java to translate web service into java code which is used to call the workflow.the third layer is the data mining service layer, there are a lot of available data mining services, and any new data mining services can be added to the workflow platform which improves the scalability of the workflow. In order to adapt to the workflow, there are a litter change to the data mining web services. The finding nodes function is added to make it more suitable for workflow.There are some characteristics on this workflow: first, it is unique, data mining by workflow in the grid data mining platform is unique, and it greatly upgrade the data mining capability. Second, the workflow is designed on standard, the workflow platform developed by using standard workflow languages, which is easy to transplant and expand.