HEC: Collaborative Research: Techniques for Streaming File Systems and Databases

许多高端计算应用程序的性能受到存储系统提供数据的容量的限制。随着处理器速度的提高，I/O性能继续显著滞后。因此，I/O很可能仍然是高端计算的关键瓶颈。在这个研究项目中，PI建议构建一个流B-树的原型，用于部署在数据库或文件系统中。该原型将实现快速插入、快速范围查询和平台无关性。PI将同时致力于如何处理不同大小的密钥，如何支持事务，如何扩展到多个磁盘和进程，以及如何为缓存遗忘和内存映射的海量数据提供O/S支持。这项工作代表了如何处理海量数据和克服传统I/O瓶颈的一个很有前途的新方向。

The performance of many high-end computing applications is limited by the capacity of memory systems to deliver data. As processor speeds increase, I/O performance continues to lag substantially. Thus, I/O is likely to remain a critical bottleneck for high-end computing. In this research project The PIs propose to build a prototype of a streaming B-tree for deployment in a database or file system. This prototype will realize t fast insertions, fast range queries, and platform independence. The PIs will simultaneously work on how to deal with different-sized keys, how to support transactions, how to scale to multiple disks and processes, and how to provide O/S support for cache-obliviousness and memory-mapped massive data. The proposed work represents a promising new direction in how to manipulate massive data and overcome classic I/O bottlenecks.