HECURA: Concurrent I/O Management for Cluster-based Parallel Storages

本项目研究系统级技术，以有效地支持基于集群的并行存储上的并发I/O工作负载。在服务器操作系统级别，研究将集中在减少并发I/O期间的磁盘寻道/旋转开销。在存储集群级，研究将集中在并发工作负载下每个并行I/O操作内的同步。对于嵌入计算任务（如并行数据聚合）的并行I/O，将研究并发负载下的存储服务器负载自适应问题，并努力将所提出的技术与服务器操作系统、基于集群的并行文件系统和I/O中间件集成。最终的软件应该对并行I/O应用程序透明，这样它们就可以在不做任何更改的情况下受益。

This project investigates system-level techniques to efficiently support concurrent I/O workloads on cluster-based parallel storages. At the server operating system level, research will focus on the reduction of disk seek/rotation overhead during concurrent I/O. At the storage cluster level, research will focus on the synchronization within each parallel I/O operation under concurrent workloads. For parallel I/O embedded with computing tasks (like parallel data aggregation), research will be performed on storage server load adaptation under concurrent workloads.Efforts will be made to integrate proposed techniques with the server operating system, clusterbased parallel file system, and I/O middleware. The resulted software should be transparent to parallel I/O applications so they can benefit without any changes.