Washington University Institute of Clinical and Translational Sciences

项目摘要/摘要 临床和翻译科学研究所与CTSA开发新的解决方案的目标保持一致 提高翻译科学的效率、质量和影响力。 家长助学金追求五个目标： 目标1：推进跨学科计划，以发展、促进、指导和留住高素质和 不同的教职员工、受训者和工作人员将科学发现转化为行动，从而成为 机构和社区变革的推动者。 目标2：促进旨在实施的研究，让不同的社区和 在翻译研究过程的所有阶段，多学科协作团队中的利益相关者， 从而展示了翻译科学的好处和影响。 目标3：整合跨越个人生命周期的研究，并将翻译科学应用于复杂和 通过有意义的研究改善个人和社区健康的多样化人口 合作和可持续的伙伴关系。 目标4：推动创新、质量、效率和纳入翻译研究，以增强 合作，并促进发现科学的实施。 目标5：应用创新的信息学和生物统计解决方案，提高每一项工作的质量和效率 转换研究阶段，并创建集成各种数据的生态系统并促进 数字资产的互操作性、使用和再利用。 本补编将侧重于以下目标： 目标1：将混合云HPC群集部署到与On-Net无缝集成的公共云 办公场所工作节点和数据存储。 目标2：在一系列HPC工作负载上运行测试，包括数据模拟、生物信息学、人工智能 智能/机器学习和成像(侧重于GPU加速器)和常规交互 工作量。我们将比较在云节点和内部部署节点上运行的工作负载的性能， 利用作业运行时间和成本作为主要指标。 目标3：评估将HPC群集和节点部署到公共云的可行性。我们将完成 这是通过检查运行目标2中描述的工作负载的成本，并将它们与 传统内部部署方法的预计成本。 补充工作的预期结果是一项全面的理解，以确定 在混合云HPC环境中，各种HPC工作负载比 传统的内部部署。

PROJECT SUMMARY/ABSTRACT The Institute of Clinical and Translational Sciences aligns with the CTSA goals of developing novel solutions to improve the efficiency, quality, and impact of translational science. The parent grant pursues five aims: Aim 1: Advance interdisciplinary programs to develop, promote, mentor, and retain highly qualified and diverse faculty, trainees, and staff who translate scientific discoveries into action, thereby becoming agents for change in their institutions and communities. Aim 2: Facilitate research designed for implementation by engaging diverse communities and stakeholders in multidisciplinary collaborative teams at all stages of the translational research process, thereby demonstrating the benefit and impact of translational science. Aim 3: Integrate research across individual lifespans and apply translational science in complex and diverse populations to improve individual and community health through meaningful research collaborations and sustainable partnerships. Aim 4: Drive innovation, quality, efficiency, and inclusion in translational research to enhance collaborations and catalyze the implementation of discovery science. Aim 5: Apply innovative informatics and biostatistics solutions to improve quality and efficiency at every stage of translational research, and create an ecosystem that integrates diverse data and facilitates the interoperability, use, and reuse of digital assets. This supplement will focus on the following aims: Aim 1: Deploy a hybrid-cloud HPC cluster to the public cloud that is seamlessly integrated with on- premises worker nodes and data storage. Aim 2: Run tests on a range of HPC workloads, including Data Simulations, Bioinformatics, Artificial Intelligence/Machine Learning & Imaging (GPU accelerator-focused), and general interactive workloads. We will compare the performance of workloads running on cloud and on-premises nodes, utilizing job run time and cost as the primary metrics. Aim 3: Evaluate the viability of deploying HPC clusters and nodes to the public cloud. We will complete this by examining the cost of running the workloads described in Aim 2 and comparing them to the projected costs of traditional on-premises approaches. The expected outcome of the supplement work is a comprehensive understanding to determine either what kinds of HPC workloads are more cost effective, or more performant in a hybrid-cloud HPC environment than a traditional on-premises deployment.