Replace Oracle with Open Source DB?
Date: Sun, 15 Mar 2009 23:09:53 -0400
Message-ID: <gpkfu0$ca2$1_at_reader1.panix.com>
BACKGROUND
[Quoted] [Quoted] We have an Oracle 10g installation. There seems to be one important application implemented in Oracle; a kind of Content Management System to control information for eventual publication on N or more different web sites.
Information professionals arrange materials into a hierarchy, add descriptive metadata (The sort of information one might find in Library Catalogs), declare the material "ready for prime time" and assign materials to the different web sites. They also control the production of images (digital pictures mostly) from the "real world" objects; but Oracle does not store the actual images themselves.
Some of the materials are "born digital" and dont need digitization control.
Data entry (descriptive metadata and arrangement) is entered into the Oracle schema holding the actual data via a web front end written in pl/sql using Apache and the pl/sql gateway. THe application is in a separate schema from the production data but has "rights" to read the tables and to execute packaged procedures living in the production schema for insert/update etc.
Independant applications running JVM's "Extract" the data by making SQL calls into Oracle, format the data into XML files, and also index the information with Lucene.
The XML files and the Lucene indexes are sent off to web farms.
The actual bit streams (pictures, sound files, etc) that are the "content" are stored outside of the Oracle system in a set of "image" servers, where "image" can be taken to be a visual image, a sound file, or any sort of media file.
The actual production data exclusive of indexes and other support structures isnt more than 2 or 3 GBytes worth.
The pl/sql code to implement the data entry application is maybe 60 KSLOC of pl/sql.
As a practical matter there are never more than a small handful of (say half a dozen tops?) simultatneous data entry operators simultaneously active.
Adding everthing togather there's less than 4 or 5 GBytes of data and 70KSLOC of pl/sql code in the whole ORacle portion system.
The heavy lifting, 24x7 web availability to consumers around the globe, is borne by the web farms and the image delivery mechanis, all outside of and only loosely coupled to Oracle through the extraction system.
I have no idea what the Oracle license fees are like. The only possibly relevant fact being that we are a non profit public institution.
THE QUESTION (at last)
The institution now favors using open source products wherever it makes
sense to do so. The decision to go Oracle for this particular project
was taken in or around the year 2000.
There are many more choices now and some options, such as mySql, which were "weak tea" then, have since become "contenders".
The situation is complicated by lack of information on my part. I am given to believe that some departments plan to acquire XML database tools, but I dont know which ones.
I can imagine that storing, retreiving, parsing, querying and ransforming XML documents will be increasingly important in the future.
With Oracle,we have XML DB, the xml data type and a whole panoply of XML integration tools. If we were to dump Oracle, and go completely "open source", (lets say that means mySql , but that's not certain either) I'm not sure how , if at all, we could recreate that easy access between the RDBMS and the XML incarnations of the information.
Does it make sense to port the application to "something" else.
If so why so, if not why not?
What options does the group think are viable?
How would we recreate the beauty of the XNL DB features in mySql or berkely DB or INNODB or whatever? Received on Mon Mar 16 2009 - 04:09:53 CET