Abstract: The problems on the rapid replica location and consistent replica access on a data-grid were studied. By analyzing the associated issues on replica caching, cached replica tracing and replica adjacency perceiving in data-grids, the paper proposes a distributed "Indexed-Cache Framework Over Data-Grids" (IFOG) and accomplishes an algorithm of the rapid replica allocation and access. The experiment shows that IFOG is good at rapid allocation of near replicas, which greatly improves the efficiency of the Replica Location Service (RLS) on data-grids. The consistent replica access algorithm over IFOG can provide better replica access speed for data-grid applications without increasing the price of replica consistency maintenance.

Key words: data-grids, cached replicas, replica allocation and access, distributed indexed-cache framework, replica consistency maintenance

摘要： 研究数据网格中快速副本定位和一致性存取问题。基于对数据网格中的副本缓冲、缓冲副本的跟踪、临近副本的感知等问题的关联分析，提出了一种数据网格上的分布式缓冲索引架构(IFOG)，并利用该架构有效地实现了快速副本定位和存取算法。模拟试验表明，IFOG架构在快速定位临近副本方面表现出色，极大地改善了数据网格的副本定位服务(RLS)的效率。同时，基于IFOG架构的一致性存取算法能够在不增加副本一致性维护代价的前提下为数据网格应用提供更高的数据访问速度。

关键词: 数据网格, 缓冲副本, 副本定位和存取, 分布式缓冲索引架构, 副本一致性维护