FC-SAN中的数据放置和访问路径选择的代价模型

网络化存储通过引入网络的概念将存储独立于服务器甚至通信网络,已经成为传统存储方式的有力替代者.然而,FC-SAN虚拟存储方式的存储性能依赖于存储对象的某些属性,在某些情况下,其性能甚至不如传统的LAN数据共享方式.就FC-SAN虚拟存储方式中的数据放置和访问路径选择对这一问题进行了研究.首先通过分析虚拟存储原理提出了一个数据访问耗时的线性模型;然后,就数据放置和访问路径选择提出了一个决策方法;并在进一步探讨这一方法的过程中，定义了“虚拟存储代价当量”的概念，用以评价FC-SAN虚拟存储环境中的数据放置的代价，从而为评价以及如何选择数据放置和访问路径提供了一种定量的手段.最后，在数字图书馆的一个海量存储原型系统中对上述的理论分析、各种条件进行了实验验证，并结合实际给出了“虚拟存储代价当量”的计算方法，验证了所提出的方法的有效性.

A Cost Model for Data Placement and Access Path Selection Problem in FC-SAN

Network storage methods (e.g. FC-SAN) with virtual storage technology are becoming a powerful substitution of DAS in digital libraries and other massive storage applications. However the efficiency of FC-SAN virtual storage strongly depends on some attributes of the stored documents. In some cases, FC-SAN may even perform no better than the sharing data on LAN. This paper illustrates this point by a study on the data placement and access path selection issues in a network storage environment. The paper first presents a linear timeconsuming model of data access through the analysis of the virtual storage principle and then gives a decision-making method for data placement and access path selection problem. In the development of this method, the concept of equivalent of virtual storage cost is defined to evaluate the data placement cost in the FC-SAN virtual storage environment. Finally, the theoretical analysis, methods, and assumptions are proved by the experimental results based on a massive storage subsystem in a digital library prototype.