Forrester研究报告：Information Fabric——企业数据虚拟化(Part VIII、摘要翻译)

The Centrally Managed Distributed Metadata Repository (DMR)
集中式管理的分布式元数据存储(DMR)

The greatest challenge with distributed data repositories is integration, especially when dealing with large amounts of disparate data and integrating them in real time. 分布式数据存储的最大挑战是如何集成、尤其是实时集成并处理大量异类数据。 Unlike EII, ETL, and data replication approaches to data integration in which each data access has to be predefined with a DMR, distributed data is integrated at runtime and requires no predefined access paths, thus providing greater flexibility (see Figure 3). 与EII、ETL和数据复制方法不同，数据集成的每个数据访问都需要通过预定义的DMR，运行时的分布式数据集成无需预定义的访问路径，灵活性更高(见下图)。 Unfortunately, the industry still lacks comprehensive standards or products to tie various metadata together. 不幸的是，目前业界缺乏完整的标准或产品来将不同的元数据结合在一起。

How is a centrally managed DMR different than EII? 集中式管理的DMR与EII都有什么区别呢？EII is a broad term that covers a collection of technologies and best practices for providing custom views into multiple data sources to integrate data and content for real-time read and write access by applications. EII是一个更宽泛的概念，它包含了应用可进行实时读写访问的多数据源自定义视图的技术集和最佳实践。 EII integrates data via views, fetching data in real time as needed instead of requiring an involved data movement process like ETL. EII通过视图集成数据、按需实时获取数据，无需像ETL那样的数据搬运过程。 Unlike information fabric, EII is mostly focused on structured data via SQL relational access and therefore usually can’t include content. EII与Information Fabric不同，EII主要专注于通过SQL对结构化的数据进行关系型访问，因此通常不包含内容。

The EII approach to integration can be a slow process, because it often requires EII platforms to communicate with multiple source systems, fetch multiple result sets, and merge them into a single result that the application receives. 使用EII来集成可能是缓慢的过程，因其通常需要EII平台与多个源系统通信、查询多个结果集并将他们整合至应用可获得的统一结果中。 With information fabric, data is represented by virtual metadata constructs, and data access can be constructed and altered in real time to cope with evolving business requirements. 使用Information Fabric，数据通过虚拟元数据结构代表，对数据的访问可进行实时修改和构建，以满足不断上升的业务需求。 The DMR component integrates data in real time and caches data to optimize data access. DMR组件将数据实时集成并使用数据缓存来优化数据访问。

The key characteristics of a DMR include:
DMR的主要特点有：

• Flexible repository architecture. 灵活的存储架构。The DMR should be able to adapt to new applications, and changing business requirements — meaning that its metamodel should be extensible, should integrate with other sources of metadata where possible, and should provide tooling as needed to configure the fabric. DMR需要适应新的应用和变化的业务需求——这就意味着元数据模型必须是可扩展的，能与其他可能的元数据源集成，同时也需要提供配置fabric的工具。

• Distribution across servers. 跨服务器的分布。 Metadata must be stored on all servers participating in the information fabric and synchronized in real time across servers. 需要将元数据存储在Information Fabric的各个服务器中，并可跨服务器进行实时同步。

• Description of the data access path. 描述数据访问路径。 The DMR must be able to describe not only the data model but also the data access path, i.e., mapping to the caches and servers where data can be retrieved. DMR必须能够描述元数据模型与数据访问路径，例如，映射数据缓存和服务器。

• Ability to span applications. 跨应用能力。 The DMR is not focused on a single application or usage context, such as customer data, but spans all types of custom and packaged applications and master data. DMR不止关心一个应用或上下相关的应用(例如客户数据)，还需要扩展到所有类型的定制/打包应用与主数据。

原文： Information Fabric: Enterprise Data Virtualization下载