专栏首页数据饕餮数据仓库专题(23):总线矩阵的另类应用-Drill Down into a More Detailed Bus Matrix

数据仓库专题(23):总线矩阵的另类应用-Drill Down into a More Detailed Bus Matrix

一、前言

Many of you are already familiar with the data warehouse bus architecture and matrix given their central role in building architected data marts. The corresponding bus matrix identifies the key business processes of an organization, along with their associated dimensions. Business processes (typically corresponding to major source systems) are listed as matrix rows, while dimensions appear as matrix columns. The cells of the matrix are then marked to indicate which dimensions apply to which processes.

In a single document, the data warehouse team has a tool for planning the overall data warehouse, identifying the shared dimensions across the enterprise, coordinating the efforts of separate implementation teams, and communicating the importance of shared dimensions throughout the organization. We firmly believe drafting a bus matrix is one of the key initial tasks to be completed by every data warehouse team after soliciting the business’ requirements.

二、面临问题

While the matrix provides a high-level overview of the data warehouse presentation layer “puzzle pieces” and their ultimate linkages, it is often helpful to provide more detail as each matrix row is implemented. Multiple fact tables often result from a single business process. Perhaps there’s a need to view business results in a combination of transaction, periodic snapshot or accumulating snapshot perspectives. Alternatively, multiple fact tables are often required to represent atomic versus more summarized information or to support richer analysis in a heterogeneous product environment.

三、解决方案

We can alter the matrix’s “grain” or level of detail so that each row represents a single fact table (or cube) related to a business process. Once we’ve specified the individual fact table, we can supplement the matrix with columns to indicate the fact table’s granularity and corresponding facts (actual, calculated or implied). Rather than merely marking the dimensions that apply to each fact table, we can indicate the dimensions’ level of detail (such as brand or category, as appropriate, within the product dimension column).

 四、总结

The resulting embellished matrix provides a roadmap to the families of fact tables in your data warehouse. While many of us are naturally predisposed to dense details, we suggest you begin with the more simplistic, high-level matrix and then drill-down into the details as each business process is implemented. Finally, for those of you with an existing data warehouse, the detailed matrix is often a useful tool to document the “as is” status of a more mature warehouse environment.

本文参与腾讯云自媒体分享计划,欢迎正在阅读的你也加入,一起分享。

我来说两句

0 条评论
登录 后参与评论

相关文章

  • AIX6.1 线程模型说明

    引文:线程模型(Threading Model)默认从进程域 (M:N 模型 ) 改为系统全局域 (1:1 模型 )

    数据饕餮
  • JDK7并行计算框架介绍一 Fork/Join概述(官方原版-英文)

    New in the Java SE 7 release, the fork/join framework is an implementation of th...

    数据饕餮
  • 数据仓库专题(21):Kimball总线矩阵说明-官方版

    Over the years, I have found that a matrix depiction of the data warehouse plan...

    数据饕餮
  • 参数化与人工智能,从计算机辅助到计算机决策,同济大学DigitalFuture演讲记录

    这是他在同济大学DigitalFuture演讲稿,为我们介绍了人工智能在建筑领域的应用。欢迎大家关注他的公众号(见文末)

    mixlab
  • Tuning a random forest model调试随机森林模型

    In the previous recipe, we reviewed how to use the random forest classifier. In ...

    到不了的都叫做远方
  • Golang memory model

    Introduction The Go memory model specifies the conditions under which reads of a...

    李海彬
  • 超弱局部间断伽辽金法研究具有高阶空间导数的偏微分方程(CS NA)

    本文提出了一种新的不连续伽辽金方法,用于求解具有高阶空间导数的几类偏微分方程。将局部间断伽辽金法和超弱间断伽辽金法的优点结合起来。首先将具有高阶空间导数的偏微分...

    非过度曝光
  • 【量化精品】通过LSTM神经网络进行时序预测针对股票市场(附Python源码)

    阅读原文 Neural Networks these days are the “go to” thing when talking about new fad...

    量化投资与机器学习微信公众号
  • How to build Multi-Language Web Sites with ASP.NET 2.0 and VS.Net 2005

    Introduction: In order to reach international markets through the Internet, sup...

    张善友
  • CodeForces 157A Game Outcome

    A. Game Outcome time limit per test 2 seconds memory limit per test 256 me...

    ShenduCC

扫码关注云+社区

领取腾讯云代金券