Systems Support for Scalable Requestt-Specific File Service

本项目提出构建一个实验系统，可以提供可扩展的特定请求文件服务。随着应用程序需求多样性的增加（持续媒体检索的确定性保证、交互式应用程序的短延迟和常规文件检索的吞吐量需求），对文件和web服务器的需求也在不断增加。新的应用程序（多媒体和数据库决策支持）对大量I/O带宽的需求也提示了提供文件服务的新方向。这些应用趋势，加上硬件（如网络连接磁盘）的可用性和更便宜的处理能力的趋势，促使我们研究以下问题：(a)如何在文件/存储系统中提供灵活的服务？（二）如何为多类型请求的综合服务提供调度和资源分配支持？（三）如何为靠近数据的迁移服务提供支援？（四）灵活服务应提供何种应用界面？本项目提出基于pc机集群构建和实验原型。原型将有助于评估所提议的特定于文件的请求服务的有效性和有效性。

This project proposed to build an experimental system that can provide scalable request-specific file service. The demands on the file and web servers continue to increase along with the increase in the diversity of application requirements (deterministic guarantees for continuous media retrieval, short latencies for interactive applications and throughput requirements of regular file retrieval). The new applications' (multimedia and database decision support) requirements of large amounts of I/O bandwidth also prompt new directions in providing file service. These application trends, coupled with the trends in the availability of hardware such as network-attached disks, and cheaper processing power, prompt us to investigate the following: (a) How to provide flexible service in a file/storage system?(b) How to provide scheduling and resource allocation support for integrated service of multiple types of requests?(c) How to provide support for migrating service close to the data?(d) What kind of application interface should be provided for flexible service?This project proposed to build and experimental prototype based on a cluster of PCs. The prototype will be instrumental in evaluating the validity and efficacy of the proposed file-specific request service.