eam@soleil.UUCP (Ed A. Mills) (08/18/89)
I am currently creating an application in VMS-5.1 (VAX) which manages a large
set of engineering specifications. I am using RMS Record Services to man-
age records which have 1 to 20 or so KEYS. I have no reason to ever sort
the data arbitrarily; that is I have used the KEYS on fields that I need to
sort by, and other fields I don't sort on. I use System services from Pascal,
or QIOWs, or standard Pascal predeclared's. Files are opened INDEXED and
KEYED, and con viently, are opened only once and never closed until the
program is done executing.
The database will have 1-100 concurrent users, and has about 20 datafiles.
Record Locks / contention are managed by an algorithm which checks file
locks in advance, then waits and rechecks n times until the file is rel-
eased or not released. If not released, the user can try later, but in
practice this condition is very rare and usually a software problem.
Record deadlocks are by design impossible, as records are immediately
used and the record released before another file is accessed.
So my question is (finally) what would a database but me that I don't
already have? I can think of several disadvantages in using a database,
namely:
1. more $$ for the software
2. necessity to learn the QL or interchange language and calls
3. slower record access and program execution through layered product
4. further decrease in program portability (now it runs on any VAX).
5. problems in keeping the database current with (everchanging) VMS
6. increased storage requirement for table entries I'll never use
I can think of a few advantages, mostly for special circumstances:
1. database conversion to a new record structure would be much easier
(right now I have to write a program to do this)
2. less prone to file corruption, and ability to correct corrupt records
3. use of tools and translators that come with the database
4. some added flexibility in that I can use table entries instead of
adding new KEYS to the records.
None of these advantages make me want to go out and spend 50K for a database.
I can do a conversion in a few hours; I back up daily; I have other tools;
and I don't expect to need new KEYS very often.
Maybe I'm missing something? I would welcome all E-Mail Pro and Con with
the RMS approach vis-a-vis the database approach.
Ed Mills
Harris Semiconductor
Melbourne, FL