Research Of Instance-Intensive Workflow Algorithm In Agricultural Data Cloud Platform

With the rapidly development of agricultural technology, the agricultural data cloud platform is emerged to adapt to the new characteristics of agricultural science data, such as big data volume and variety, wide data distribution, real-time data processing. The project of Software Platform of Process and Convergence National Agricultural Science Data provides auditing services for the agricultural big data. Although the data auditing is composed of simple business processing procedures, it needs to cope with the highly concurrent user request with the Qo S(Quality of Service) requirements, which lead it to be align to instance-intensive workflows. However, the Qo S parameter expression model and the workflow scheduling algorithm, specific to the agricultural data auditing, are not covered in the existing instance-intensive workflow research and engineering work. Work in this thesis is aiming to filling in this blank in the workflows research. The main contributions are as the follows:1) This thesis studied the architecture of instance-intensive workflows on agricultural data cloud platform. According to the workflow reference model and cloud computing system architecture, this thesis described how to build workflow system on cloud environment and how to build the architecture of an instance-intensive workflows in agricultural data cloud platform. At last, this thesis explain the operating mode with an example of business process in the background project.2) This thesis determined collection of Qo S parameters. First, through the analysis of agricultural science data gathering mode and its own characteristics, instanceintensive workflows platform features and the requirement of quality of service of users, this thesis chose Qo S parameters to describe preferences of users. Next, this thesis defined Qo S expression model and numerical constraint to support the calculation in the algorithm. Finally, this thesis defined a lightweight Qo S parameters retrieval mechanism for users to set the Qo S parameters;3) This thesis proposed a Multiple-Qo S-based Instance-intensive Scheduling Algorithm in Agricultural Cloud Environment, MQISA. First, this algorithm used the utility function on Qo S was calculate Qo S parameters to optimize the sequence of scheduled instances and to make the algorithm flexible. Second, overall deadlines was decomposed into sub-deadlines to assign to every task. At last, different task allocation strategy was performed by different preferences of users.Based on the above research results, this thesis designed two sets of experiments of MQISA algorithm on the simulation platform Workflow Sim. The result of first experiment, which set the same premise with CTC algorithm, showed that the average execution cost of MQISA reduced 17.65% comparing with CTC, which proved the effectiveness of Qo S parameters choosing. The result of second experiment, which used three preferences types to describe three typical tendencies of users, showed that MQISA can support the different users’ preferences of execution time and execution cost.