In response to Rob Armstrongıs letter in the March 1998 DBMS Letters Column:
We are successfully using Ralph Kimballıs dimensional modeling approach for the data warehouse in a large bank in Mexico. We have a high-availability cluster with Oracle, and we use decision-support tools that need dimensional models rather than entity-relationship (ER) models to function.
Oracle (the biggest database vendor, in my opinion) is moving to the dimensional approach: Its new optimizer includes star queries, bitmapped indexes, hash join processing, and other features specially designed to exploit star schemas. My company is using all these features and we (and our users) are very happy with the results. Response times are very fast, and we can perform an in-depth analysis on the information.
With ER modeling, the users themselves must navigate the model (with or without the metadata) to get answers to their business questions. With dimensional modeling you can easily use tools to navigate all your models. We donıt need to make any queries to answer user questions because the tools can handle the dimensional models. (We have at least 15 conformed fact tables and 25 conformed dimension tables. We track the history of 400 branches throughout Mexico and we plan to store three years of information online. We will have about 1.5TB of information in three years.) I think that ER modeling is only useful for the transactional operation of the business, and only the IS people can truly understand it. Imagine the need to completely understand the SAP database (12,000 relational tables) or the Oracle Financials database (8,000 tables). I donıt think everyone can, or should, do that.
CASE tools that make data models (such as ERwin 3.5) use Kimballıs dimensional modeling approach and provide very nice tools for data warehouse modeling and documentation.
I think you need to open your mind to the new ideas. I feel that everybody in the data warehousing business needs to forget the older theories from Dr. Codd about normalized databases to get a data warehouse functional and working. I urge you to read Kimballıs book, The Data Warehouse Toolkit, to get a better understanding of the theory of architected data marts forming a complete data warehouse. You will be very happy with the results.
Xavier Angel Montalvo Alba
Grupo Financiero Banorte
xamontalvoa@gfnorte.com.mx
I just read Tom Spitzerıs February 1998 DBMS column, "Tiers Without Tears." I think itıs a great article, particularly because it is written in simple language that makes it easy to understand, even for those who havenıt worked on three-tier architectures.
I work for Star TV in India, and we are developing a three-tier application. We are using Visual Basic (VB) 5 to build our front and middle tiers. We have just implemented it using DCOM and are now moving toward MTS. At first I was against the move toward MTS, mainly because of my apprehension that MTS is not stable and not many people use it. After reading this article, however, I was convinced that implementing MTS in our application is indeed a smart move. I would again like to compliment Tom for his presentation of the three-tier technology in such easy terms.
I believe that in order to implement MTS through our application we will have to use a VB-specific MTS API. I have studied that to an extent. Unfortunately, there arenıt any sample applications available to suggest those APIs, other than the bank application that comes with MTS 1.0, which we have installed. Any ideas where I could find some sample applications?
Kapil Israni
k_israni@hotmail.com
Thanks for your kind comments. I must have been much more persuasive than I intended because I was not trying to convince people that MTS is ready for "prime time." Even version 2.0 lacks some functionality that I would like to see. However, the programming and deployment model is so productive that it compensates for the relative lack of scalability and robustness of the NT platform.
You may also wish to read my article on MTS 2.0 in the April issue of DBMS. You should review the updated bank sample code that comes with MTS 2.0. MTS 2.0 is available to Windows NT licensees as a free download. There is extensive documentation as well.
ı Tom Spitzer
Thank you for Clara Parkesı article, "Whatıs the beef behind the OLAP blab?" (January 1998). I am the senior investment adviser for a retail business in Taiwan. I was recently appointed the co-head of our IT restructuring plant. Not coming from a technical background, I had to take a crash course. I went to the Web sites of all the relevant vendors and downloaded their white papers. Guess what? They just made me more confused. I am now reading David Linthicumıs book Guide to Client/Server & Intranet Development, and it is finally becoming clear to me.
Thanks for publicly speaking out about all the nonsense mumbo jumbo that some of the press ı and most of the companies ı put out there. It only makes our lives miserable. And, as you point out, they often sound very distant (sometimes stupidly so) from what they are trying to describe. Keep up the good work; Iıve become a loyal reader of DBMS and your column.
Jim Wang
ub40@ar.ar.com.au
I read with great interest Maurice Frankıs insightful article, "Death of the Database?" (March 1998). I agree with his question: "If the RDBMS is dead, why are DBAs and developers still getting away with murder?" However, I would like to examine the point made in relation to packaged applications. As we know, SAP is considered a leader in this market, and SAP R/3 does not require an RDBMS. If the user is willing to forgo the numerous benefits of having a real RDBMS, they are free to use SAPıs proprietary database. It would be interesting to find out what percentage of current SAP R/3 implementations have gone this route. I am also wondering if information technology outsourcing (ITO) and business process outsourcing (BPO) can be viewed as contributors to the so-called "Death of the Database." Undoubtedly, BPO solutions that claim to deliver human resources, payroll, financial services, purchasing, and order entry do so by having a strong DBMS. Is it possible that RDBMSs appear to be losing momentum because they are being disguised as BPO?
Arshak Navruzyan
Net Computing Inc.
arshak@netcomputing.com
Thank you very much for your comments. You raise some good questions about SAP and outsourcing. I am not aware of any studies that have tried to identify the impact of these options. With outsourcing, if an application requires a DBMS, then the location and owner of it might be transferred, but the need doesnıt go away.
ı Maurice Frank
I am a consultant working with Peoplesoft ERP, and David Linthicumıs Application Architect column in the January 1998 issue of DBMS is one of the best articles I have read on the entire enterprise in a long time. I share Davidıs belief that "supply to consumption" is the way ERP will eventually go, but I feel there are a number of barriers holding it back at present.
In my experience, clients are currently scrambling to solve their Year 2000 problems, using up their budgets maintaining Cobol or migrating to ERP, PowerBuilder, and so on.
Resources, including the number of people who understand SAP, Baan, and PeopleSoft, are low, driving up costs. There is also a shortage of programmers and business analysts. If we consider Tuxedo CrossRoads and other midtier and cross-integration products, then the shortage is even worse. Companies that do have good people tend to lose them to consulting companies who can pay (and charge) more money. I hope this crazy cycle will end soon.
I see three phases of development in the ERP cycle:
Editorıs Note: In our April 1998 issue, we published a sidebar in Martin Rennhackkampıs Server Side column about JB Developmentıs product Harmonia. Unfortunately, JB Development has since been reorganized and the contact information listed in the sidebar is no longer valid. Until the reorganization is complete, Harmonia will not be available. JB Development does plan to offer several embedded products based on its high-performance row-access engine.