Efficient use of commercial cloud for astronomical data archive and data analysis

The amount of astronomical data that needs to be archived, calibrated, and processed continues to increase as telescopes and observing instruments advance. Securing necessary resources to store and process ever-increasing data is an operational challenge. To solve these issues, we conducted a demonstration experiment using ALMA archived data to efficiently utilize a commercial cloud for archive storage and data analysis pipeline processing. In archiving, a hybrid configuration combining on-premise storage and cloud based short-term and long-term storages is cost-effective, considering the trends on the number of data downloads over time since the data was obtained. In the data analysis processing, information on processing time and resource usage, such as memory and CPU core, measured during the pipeline process of approximately 400 observation data sets was analyzed, and a model was created to estimate processing time and the required amount of resources from the observation parameters. Based on the model created, the amount of required resources is predicted based on observation parameters, and an instance with the necessary and sufficient resources for pipeline processing is launched on demand on the cloud. These pipeline processes were completed with resources in a processing time comparable to that of on-premise ones. Since prices, services, computing resources, etc. on commercial cloud are updated frequently, we plan to continue making periodic estimates.