作者： 來(lái)源： http://www.csai.cn 2006年9月13日
DATA WAREHOUSE
It is often said，the age of the Industrial Revolution has finally been completed and the world has entered the age of the Information Technology revolution（see Fig. 7-5）．It is our belief that the need for data warehouse applications is one of the manifestations of this Information Technology age．A data warehouse is becoming more of a necessity than an accessory for a progressive，competitive，and focused organization[1]．It provides the right foundation for building decision support and executive information system tools that are often built to measure and provide a feel for how well an organization is progressing toward its goals[2]．
1．COLLECTING OPERATIONAL DATA
Advances in computer and networking technology have led to the introduction of very powerful hardware and software platforms that can collect，manage，and distribute large amounts of pertinent data．In the case of a business application，detailed transactions are often generated during product-or service-related interactions．These transactions are not limited to commercial sectors．They are also found in sectors such as，government，health care，insurance，manufacturing，finance，distribution，education，and so on．Any enterprise that has some computerized record keeping systems and is interested in deducting or drawing logical conclusions from their voluminous，granular，and detailed information pool should consider building an enterprise-level data warehouse application[3]．These enterprises will then be capable of improving their insights into the trends in their operations and eventually increase the accuracy of their forecasts and plans．The effectiveness of the data warehouse application intensifies especially when the operational data resides in distributed，non-homogenous systems and replace manual data gathering and reconciliation procedures．
Operational data is the highly structured sets of information that support the ongoing and day-to-day operation of an organization．In case of a decentralized organization，operational data is generated at remote locations sometimes in non-homogeneous distributed systems（see Fig. 7-6）．Distributed systems can span many different geographical locations and time zones．They are configured to provide scalability，visibility，and tracking capabilities of business processes．For instance，the order is entered by a customer representative in one site．The financial state of the order is verified at another site．Once approved，it is forwarded to manufacturing to be assembled[4]．Finally，the shipping staff is alerted to fulfill the order that was booked at the remote site[5]．Standard reports or ad hoc queries that inquire about the details of these events are typical examples of operational reports．They are generated on a regular basis．Any delay in their processing will have a significant disruption to the normal operation of that business．
2．CONVERGENCE OF MANY COMPUTER TECHNOLOGIES
The infrastructure that supports the data warehouse application relies on the same technologies that most other applications are dependent upon．The difference is in the variety and specialization at the product level that can greatly improve the quality of the data warehouse infrastructure．
Below are some technologies that have made their mark in the data warehouse marketplace．In order to produce a data warehouse that best meets users needs，these underlining technologies have to be evaluated as part of the periodic resource capacity planning．Depending upon the requirements and resources available，the best combination can be selected and configured．
• Server technology．
• Client technology.
• Database Management System（DBMS）technology.
• Networking technology．
． Mass storage technology.
• Data presentation and publication requirements.
• Software engineering methodology and tools．
COMMON CHARACTERISTICS OF A DATA WAREHOUSE
（1）Data is divided into three categories．
（a）Reference and Transaction Data．
• Includes lists，charts，and transaction data from source systems．
• Originally generated in the source systems．
• Can be kept in the data warehouse or an operational data store system．
• Is loaded into the data warehouse on a regular basis．
• Should never change once in the data warehouse（data correction and refresh are exceptions）．
• May be purged from the source．
• Is archived in the data warehouse if purged from the source．
（b）Derived Data．
• Is based on the reference data and certain business rules．
• Can always be re-created．
• Business rules must be approved by end-users．
（c）Denormalized Data．
• Is based on the detailed reference data．
• Is prepared periodically．
• Is the foundation for OLAP tools．
（2）Enhancements are done in an iterative approach．
（3）Enhancements should be based on the overall architecture．
（4）One end-user tool may not be adequate for all analytical needs．Depending on the amount of data and type of queries，different end-user tools must be selected．
（5）Transaction-level database recovery is not necessary．
（6）Data warehouse platform should he tuned for performance rather than quick recovery purposes．
NOTES
[1] 此處accessory是指附屬機構，附庸。
[2] that后面引出的是定語(yǔ)從句，是對前面的information system tools的說(shuō)明，而provide a feel for后面的how…是一個(gè)賓語(yǔ)從句。
[3] 長(cháng)句，主句為Any enterprise…should consider…主語(yǔ)Any enterprise后面是由that引出的有并列謂語(yǔ)的定語(yǔ)從句that has… and is interested in…
[4] manufacturing制造（業(yè)）。
[5] that引出的是定語(yǔ)從句，修飾order。
KEYWORDS
data warehouse 數據倉庫
information technology（IT） 信息技術(shù)
decision support 決策支持
operational data 操作數據
platform 平臺
transaction 事務(wù)（處理）
distrbuted system 分布式系統
infrastrticture 基礎設施
client 客戶(hù)
mass storage 大容量存儲器、海量存儲器
data refresh 數據刷新
information pool 信息庫
iterative approach 迭代方法
database recovery 數據庫恢復
翻譯:
人們常說(shuō)工業(yè)革命時(shí)代已最終完成，世界進(jìn)入了信息技術(shù)革命時(shí)代（參見(jiàn)圖7-5）。我們相信數據倉庫應用的需求就是信息技術(shù)時(shí)代的標志之一。數據倉庫對于不斷進(jìn)取的、具有競爭力的、成為關(guān)注焦點(diǎn)的組織來(lái)說(shuō)不只是一個(gè)附屬物，而是不可缺少的。它提供建立決策支持和執行信息系統工具的基礎，這些常常是用于衡量和評價(jià)某個(gè)單位是否正在向預定目標前進(jìn)的標志。
1．操作數據的收集
計算機和網(wǎng)絡(luò )技術(shù)的進(jìn)步已經(jīng)導致功能非常強大的，能收集、管理和分發(fā)大量有關(guān)數據的硬件和軟件平臺出現。在商業(yè)應用中，與產(chǎn)品和服務(wù)相關(guān)的交往經(jīng)常會(huì )涉及繁瑣的事務(wù)處理，這些事務(wù)不限于商業(yè)方面，它們也會(huì )在政府、保健、保險、制造、財經(jīng)、分銷(xiāo)、教育等方面出現。任何企業(yè)，如果有了某些計算機化記錄保存系統，并且很希望從大量松散詳細的信息池中演繹或作出邏輯結論，就應考慮建立一個(gè)企業(yè)級數據倉庫應用系統。之后，這些企業(yè)就能夠提高洞察運營(yíng)趨勢的能力，最終提高其預測和計劃的精確性。尤其是操作數據分布在多個(gè)異構系統上，并可取代人工數據收集和調節過(guò)程時(shí)，數據倉庫應用系統能提高效率。
操作數據是高度結構化的、支持一個(gè)單位開(kāi)展日常工作以及持續發(fā)展的信息集。如果一個(gè)單位的各部門(mén)是分散的，則操作數據有時(shí)是在遠程的分布式異構系統中產(chǎn)生的（如圖7-6所示）。分布式系統可以跨不同的地域和時(shí)區。分布式系統可以配置成使商業(yè)過(guò)程具有伸縮性、可視性和跟蹤的能力。例如，由客戶(hù)代表在一個(gè)地方輸入訂單，而訂貨人的財經(jīng)狀況在另一個(gè)地方進(jìn)行驗證。一經(jīng)批準，則訂單傳向加工廠(chǎng)裝配。最后，由于是在遠端站點(diǎn)預訂的，因而要提醒裝運人員完成訂單規定的工作。標準報告或特別詢(xún)問(wèn)是上述事件細節的查詢(xún)操作報告的典型實(shí)例，報告是正規的。這一過(guò)程中的任何延誤都會(huì )對正常的運作造成很大的損害。
2．聚集了多種計算機技術(shù)
支持數據倉庫應用的基礎設施與其他大多數應用所依靠的是相同的基礎設施不同的是產(chǎn)品層次上的種類(lèi)和專(zhuān)用性方面，這在很大程度上提高了數據倉庫基礎設施的質(zhì)量。
下面是在數據倉庫市場(chǎng)上最具代表性的一些技術(shù)為了產(chǎn)生能最好地滿(mǎn)足用戶(hù)需求的數據倉庫，作為周期性資源能力計劃的一部分須對這些重要技術(shù)進(jìn)行評價(jià)。根據需要和可用的資源，可以選擇和配置最佳組合。
服務(wù)器技術(shù)。
客戶(hù)技術(shù)。
數據庫管理系統（DBMS）技術(shù)。
聯(lián)網(wǎng)技術(shù)
海量存儲技術(shù)。
數據表示和發(fā)布需求。
軟件工程方法學(xué)和工具。
3．數據倉庫的一般特性
（1）數據分為3種類(lèi)型。
①基準和事務(wù)數據。
包括出自源系統的列表、圖表和事務(wù)數據。
源系統產(chǎn)生的原始數據。
可以保存在數據倉庫或在操作數據存儲系統中的數據。
按正規格式裝入到數據倉庫的數據。
數據倉庫中從未更改的數據（數據校正或刷新除外）。
可從源系統清除的數據。
從源系統清除而又在數據倉庫中存檔的數據。
②導出數據。
基于基準數據和一定業(yè)務(wù)規則的數據。
總能再生成的數據。
必須由最終用戶(hù)認可的業(yè)務(wù)規則。
③非正規化數據。
基于詳細的基準數據的數據。
周期性準備的數據。
用于OLAP 工具的基礎數據。
（2）加強了迭代方法。
（3）增強應基于整個(gè)體系結構。
（4）一種最終用戶(hù)工具不一定適應所有的分析需求。應根據所查詢(xún)的數據量和類(lèi)型，選擇不同的終端用戶(hù)工具。
（5）不必有事務(wù)級數據庫恢復。
（6）數據倉庫平臺應以性能而不是以快速恢復為目標進(jìn)行調試。
