DBMS, July 1998
I enjoyed David Linthicumıs March 1998 article "Integrating Enterprise Applications." It struck me, however, that he didnıt address two of the issues that I have found problematic: getting base data from the legacy applications into the new package database and integrating the new package with the rest of the application portfolio.
As developers, we rarely have the luxury of installing a new application directly from a paper system; the users already have some system (with lots of data) installed, and it is impractical to re-enter that data into the new application. This is especially true when the installation strategy is "big bang" (go all at once) instead of a phased approach.
Even though a package may claim to be "enterprisewide," I have not yet seen a situation where all current systems are replaced by a new package. You therefore have a situation where accounting/general ledger is on one package, manufacturing (MRPII) applications are on another package, and human resources is on another package. Getting the data ported from one to another of these can be a major technical headache.
Barb VanEseltine
vaneselbm@aol.com
I read Tom Spitzerıs article, "All Data to Consumers" in the May 1998 issue of DBMS. This article talks about Universal Data Access as a cure-all to the problem of accessing heterogeneous data sources. I would like to remind you that ODBC was touted in a similar way. I work on IBMıs middleware solution for accessing heterogeneous data sources, DB2 Universal DataJoiner, and I feel that OLE DB may succeed in providing a uniform data interface, but will fail in performance unless the following issues are addressed:
Without these, a middleware solution that attempts to construct plans involving multiple data sources will fail miserably. Unless something significant is done about the way OLE DB exposes the functions, capabilities, and the cost model of the remote data sources, the performance will be extremely poor. In my opinion, DataJoiner, on the other hand, understands the capabilities of the remote data sources and constructs optimal plans based not only on executing data as close to the data source as possible (like ISG Navigator) but also on the capabilities of the optimizer. In fact, executing a TPC-D query through DataJoiner at a remote data source is significantly faster than executing it at the remote data source in many cases, even though data resides at the remote data source. For more information on DataJoiner, see the product Web page at: www.software.ibm.com/data/datajoiner.
Shivakumar Venkataraman
shiv@us.ibm.com
I want to congratulate Tom Spitzer on his excellent overview of Universal Data Access, ("All Data to All Consumers," DBMS, May 1998.) The article was clear, well written, and almost totally devoid of the kinds of arcane buzz-words that so often frustrate readers. My compliments on DBMS in general. I was directed to DBMS Online (www.dbmsmag.com) by an article in DevX newsletter. Iım impressed!
Robert Altman
robert.d.altman@boeing.com
Correction: In our June 1998 cover story, "Java in the Database Server" by Nelson King, the following contact information for Informix should have been included: