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

The Distributed Cache Directory (DCD)
分布式缓存目录

Disk latency is often the single largest factor in elongating application response times, especially when large amounts of random data need to be accessed quickly, or data is distributed across many servers. 磁盘延迟通常是使应用响应时间变长的最大原因，尤其是当需要快速随机访问大量数据或数据分布在许多服务器上的时候。 Although data caching can overcome disk latency, in the past it has been tied to the server holding the data, so applications still need to know which server to access. 虽然数据缓存能克服磁盘延迟的问题，但传统都是将数据绑定在服务器上，所以还是需要知道去访问哪台服务器。The DCD goes beyond basic data caching by presenting a virtualized data cache that spans multiple servers and geographical locations. DCD通过虚拟化数据展现超越了基本的数据缓存从而可跨越多个服务器和地理位置。

The DCD is not just about duplicating cached data on every server in the compute farm. DCD不只是在群集的每个服务器上复制缓存数据。Rather, it is based on intelligently managing at runtime what data is cached and where, based on actual patterns of data access and usage. 更重要的是，它根据实际数据访问与使用模式，基于运行时智能管理不同数据的缓存位置。Regardless of where the data resides, the DCD should be able to seamlessly access other data caches stored on other servers using a distributed cache registry (DCR) (see Figure 2). 不管数据驻留在何处，DCD应能够使用分布式缓存入口(DCR)来无缝访问存储在其它服务器上的数据缓存(见下图)。The DCR should contain links to cache contents from all other servers in the information fabric, enabling either direct routing of requests to the home cache or creating a synchronized cache replica when access patterns justify the cost of that extra copy. DCR应该包含Information Fabric所有其它服务器缓存内容的链接，支持请求直接路由至根缓存或当访问模式认同额外拷贝的成本后同步创建缓存的复制。The key characteristics of a DCD include: DCD的关键特性包括：

• Support for heterogeneous platforms. 异构支持平台。

• Efficient distribution across servers.跨服务器的有效分布。

• Data integrity. 数据完整性。

• Awareness of other cached data. 知晓其他缓存数据。

• Support for all kinds of information. 支持所有类型的信息。

• Optimized path to cached data. 优化数据缓存路径。 Cached data will be duplicated in some cases, and the DCD should automatically use the optimum path to retrieve data quickly from disparate servers. 缓存数据在某些情况下可能被复制，所以DCD应该自动利用最佳路径从异构服务器中快速检索数据。

原文： Information Fabric: Enterprise Data Virtualization下载