CI-ADDO-NEW: Storage System Trace Collection

该奖项是根据2009年《美国复苏和再投资法案》(Public Law 111-5)资助的。存储系统研究人员通常使用来自各种来源的I/O跟踪来推动他们的分析和研究。但是，这些I/O跟踪无法充分捕获I/O对应用程序性能的影响。此外，它们不能与整个系统的其他部分一起扩展，无论是应用程序、系统还是处理器。因此，很难重放跟踪以用于存储系统评估。这尤其有问题，因为许多现有的公开可用的I/O跟踪都相当陈旧。该项目的重点是向存储研究社区提供一组跟踪，以便准确地回放和评估存储技术。跟踪将来自存储系统层次结构中不同级别的真实系统，包括网络层、文件系统和低级磁盘。这些跟踪包括存储社区中当前不存在的跟踪，如iSCSI跟踪、虚拟机跟踪和MS Exchange跟踪。该项目还将提供一套工具，用于跟踪依赖于应用程序的I/O、重放跟踪和分析。这些工具和收集的痕迹将通过公开来源作为社区资源提供。该项目的更广泛影响包括提供工具和痕迹，帮助存储系统研究人员开发新的I/O系统技术。由于存储系统对当前以信息为基础的经济至关重要，因此在了解I/O系统行为方面的进步应该非常有益。它还应该对国家利益的数据密集型应用程序有用，如基因组计划、天气模拟和数字图书馆。

"This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5)."Storage system researchers have typically used I/O traces from various sources to drive their analysis and research. However, these I/O traces fail to adequately capture the effect of I/O on application performance. Moreover, they do not scale with other parts of the overall system whether it be application, systems, or processor. As a result, it is difficult to replay the trace for use in storage systems evaluation. This is particularly problematic since many of the extant publicly available I/O traces are quite old.The focus of the project is to provide to the storage research community a set of traces that will allow for accurate replay and evaluation of storage technologies. The traces will be from real systems at different levels of the storage system hierarchy including network layer, file system, and low-level disk. These include traces that do not currently exist in the storage community such as iSCSI traces, virtual machine traces, and MS Exchange traces. The project will also provide a set of tools for tracing application dependent I/O, replaying traces, and analysis. The tools and collected traces will be made available as a community resource through publicly available sources.The broader impacts of this project include providing tools and traces that can assist storage system researchers in developing new I/O system technologies. Since storage systems are critical to the current information-based economy, advances in understanding I/O system behavior should be highly beneficial. It should also be useful to data-intensive applications of national interest such as the genome project, weather simulation, and digital libraries.