Meteorological data layout and task scheduling in a multi-cloud environment

The meteorological cloud mainly provides computing ability and meteorological datasets for meteorological model tasks. If the location of the required dataset and the execution location of the task are different, this will consume a large amount of time and bandwidth to transfer the data for the task. Meteorological data layout allocates meteorological datasets to various clouds. Because meteorological datasets are required by multiple meteorological model tasks and multiple times, the data layout is very important in the meteorological clouds. This paper focuses on how to layout out the meteorological datasets based on the association (internal meteorological datasets and between meteorological model tasks) and schedule resources for meteorological model tasks in the meteorological cloud. First, to find the association in the meteorological datasets and meteorological models, we use Apriori algorithm to mine frequent itemsets between datasets used by different meteorological models, and then we use the result to help layout meteorological data. After that, we present a heuristic algorithm for scheduling meteorological tasks. Finally, simulation comparison shows that the meteorological data layout method has a lowest value in the number of involved clouds for every task, the average size of transmitted datasets from other clouds, and the average time of transmitted datasets between clouds. We also prove that the scheduling method based on the data layout increases the number of completed tasks before their deadlines and reduces the average execution time.