CYBER-INSIGHT: Evaluating Cyberinfrastructure Total Cost of Ownership

点击翻译按钮获取中文摘要

Academic research computing technologies is rapidly changing. In addition, new methods in gleaning insight from the deluge of data that is becoming available in science and society are fueling demand for computing infrastructure, and motivating commercial companies to provision infrastructure similar to those found at academic environments. These factors, as well as others, are putting pressure on academic institutions to find cost-effective approaches to providing computing infrastructure to maximize scientific benefit while minimizing cost. This project seeks to understand the cost-effectiveness of different methods of delivering High Performance Computing capabilities to the academic research community. Specifically, the project proposes to build, publish, and regularly update comparisons of the evolving total cost of ownership (TCO) of on premise HPC clusters and data storage systems relative to NSF-supported facilities and commercial cloud services such as Amazon AWS, Microsoft Azure, and Google Cloud Platform. The results of the project will have broad applicability to the national research community, institutional resource providers, and funding agencies. In addition, the project is collaborating with many organizations to promote long-term maintenance and data sharing of the project outcomes, with the goal of creating a sustainable infrastructure for long-term analysis.Concretely, the team proposes to build and host a Jupyter notebook that will enable the cyberinfrastructure (CI) community to add, modify, and explore total cost of ownership (TOC) models based on a variety of usage patterns and performance expectations. This will allow the project to explore questions such as when is it cost-effective to use commercial clouds compared to university clusters to service researcher needs? How do staff costs, hardware utilization, capital depreciation, replacement costs, the value of money, return on investment, power and cooling costs, and bandwidth charges affect TCO? As science gateways hide resources behind web interfaces, when and where should such queries execute? In addition to investigating these important questions, the project is also expected to facilitate community building, shared experimentation and comparisons. The ultimate goal of the project is to develop comparative models that allow funding agencies, institutional resource providers, and individual users to evaluate the costs and benefits of specific choices and subsidies. Finally, the project will collaborate with the Coalition for Academic Scientific Computation (CASC), the Association of American Universities (AAU) and the Association of Public and Land Grant Universities (APLU) senior research officers (SROs) and CIOs to promote long-term maintenance and data sharing of the project outcomes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.