Snowmass Letter of Interest - Analysis Facilities - CompF5
暂无分享,去创建一个
primary interface for analyzers, we will provide access to Jupyter notebooks “as a service” using JupyterHub. As a secondary interface, we will continue to provide shell access. Authentication and authorization will be performed using a federated identity model (subject to the security constraints of the ecosystem where the facility is sited). The need for computation on non-CPU architectures (e.g. GPUs) while still maintaining the fast “time-to-insight” required for exploratory analysis is growing as the average analysis dataset size grows. Batch job queues operate on a non-interactive timescale, whereas emerging technologies can provide sub-minute turnaround for tasks that are too large to fit on a single compute node. Moreover, a clear and approachable system for probing the structure and content of the data in real-time vastly improves the understanding and efficiency for analyzers, significantly lowering the bar for new students, and even experienced physicists switching experiments.