High Performance Integration of Advanced Tertiary Stores

必须解决处理器和I/O性能之间迅速扩大的差距。存储系统中两个耐人寻味的趋势是新技术的出现，这些技术受到读写性能不对称的影响，以及数据传输速率和访问时间之间的差距越来越大。当前的例子是具有自动库机制的第三级存储系统。这样的系统体现了由于机械介质处理而导致的数据传输速率和访问时间之间的巨大性能差距，本工作的目标是通过开发自动化第三级存储库的高性能集成技术来解决这些问题。这些问题包括：1.通过使用日志结构的写缓冲体系结构来解决读写性能不对称的问题。通过在从本地存储迁移时将同一介质卷上的相关数据预先分组，最大限度地减少机械访问延迟。这项研究将允许对海量存储进行高性能访问，从而直接受益于当今不断增长的存储需求。它解决了在数据速率远远超过访问速度的情况下减少延迟的重要问题。它还通过缓解读写性能不对称的问题，使磁光和体全息等新兴存储技术能够有效地在通用场合采用。

The rapidly growing disparity between processor and I/O performance must be addressed. Two intriguing trends in storage systems are the appearance of new technologies that suffer from asymmetric read-write performance, and the widening gap between data transfer rate and access time. A current example is a tertiary storage system with an automated library mechanism. Such a system exemplifies the wide performance gap between data transfer rate and access time due to mechanical media handling.The goal of this work is to address these issues through the development of techniques for high-performance integration of automated tertiary storage libraries. These issues include:1. Addressing the issue of read-write performance asymmetry, though the use of a log-structured, write-buffered architecture.2. The minimization of mechanical access delays, by pre-grouping related data on the same media volume, as it is migrated from local storage.This research will be of direct benefit to today's growing storage needs by allowing high-performance access to vast storage. It addresses the important issue of latency reduction in cases where data rate far exceeds access speeds. It also enables the efficient general-purpose adoption of emerging storage technologies, such as magneto-optical and volume holographic, by alleviating problems of read-write performance asymmetry.