IDMS Review / Interview Questions

Basics of IDMS – great for reviewing for interviews

IDMS is a network (CODASYL) database sold by Computer Associates (originally
marketed by Cullinane/Cullinet, the first software company to make it on the New
York Stock Exchange). 

IBM’s IMS is a hierarchical database.  DB2, Oracle, Microsoft/SQL,
Informix, etc… are relational databases.   IDMS does have some
relational capabilities, but based on what I have seen, only a small percentage
of sites have taken advantage of this option (primarily because it is an
extra-cost option from CA). 

DBA = Database Administrator  – Some sites separate the duties of the
DBA and the System Programmer.  At other sites, the DBA provides both
functions.  A system programmer is more likely to do installs and work with
exits, where as the DBA is more likely to deal with logical and physical
database design, schemas, DMCL, etc…

Click here to see Bachman (schema) diagram

CA = Computer Associates – the company that bought IDMS from Cullinet in
approx 1989.

Three types of record storage modes

   1) CALC – random/hash algorithm used to assign a DBKEY target
page to a logical key.

   2) VIA – physically clustered near other records (usually CALC
records) in order to save I/O by reading multiple records on a single IDMS page
in a single I/O.

   3) DIRECT – records stored on a suggested target page (based on
DBKEY).  Not used near as often as the first two.

Other terms and quick review items for programmers:

   1) DBKEY – a 4 byte key, usually 3-byte page number and
1-byte  line index that uniquely identifies each physically stored IDMS
record.  May also be called a "pointer" or "dbkey
pointer".

   2) AREA – a range of IDMS pages used to stored IDMS data: Three
critical factors, starting-page, number of pages (or ending page), and
page-size.

   3) FILE – an IBM data set.  Areas can be mapped to files on
a one-to-one, one-to-many, or many-to-one basis.

   4) DMCL – ties together journals, buffers, and files for an IDMS
system.  There is always one global DMCL for a CV, but batch local jobs
could have custom tailored or "local" DMCLs for performance.  The
DMCLs are usually maintained by the DBA. 

   5) CV vs Local – all batch jobs run either under CV (Central
Version) or in Local mode.  CV provides lock protection, journaling, etc…
You should backup before running a local-update job.  Report jobs are often
run in local mode (no backup needed).  Local is more efficient because it
doesn’t have the overhead of the CV.  See #29 and #30 below.

   6) TP-monitor – IDMS online programs can use any of the three
methods to communicate with a terminal.  You should ask your prospective
employer which kind of "shop" they are.

        (1) DC – IDMS has its only TP-monitor
called DC

        (2) CICS – IDMS works with IBM’s
CICS.  Either you can write CICS programs which access IDMS database and
use CICS maps to display and get the data, OR you can run IDMS under CICS/UCF
(Universal Communication Facility), and programmers and user don’t need to know
they are really running under CICS.  Programs use IDMS maps, which gets
translated into the appropriate language for CICS and terminals to
understand. 

    Also ask if the company uses ADS or ADSO (some people
leave the "O" off the end, both are commonly used).  ADS can run
under DC or CICS, but some shops might use CICS for their TP-Monitor, but then
use ADS for their programming interface.

    ADS stands for Application Development System, and is a
separate language within IDMS.  You build IDMS Maps using the MAPC compiler
(and online task), then connect the maps, work-records, code and subschemas into
Dialogs.  Code is written in IDD modules, also called Response Processes.

   7) CAS – Cullinet Application System? – a manufacturing,
accounting, payroll, inventory, PO, AR/AP system sold by CA that runs entirely
under IDMS.

   8) Schema/subschema – a schema contains the logical database
definintion.  A subschema is a subset of a schema.  Each batch program
must refer to one (and only one) subschema/schema combination.  These are
usually created and maintained by the DBA.

   9) DML = Database Manipulation Language, special IDMS verbs,
such as OBTAIN, MODIFY, STORE, ERASE, COMMIT, READY, FINISH, ROLLBACK… 
Programmers write DML code in their COBOL programs, which are run through the DMLC preprocessor to turn the special IDMS
verbs into standard COBOL calls (to the IDMS subroutine). 

  10) A many-to-many relationship – is accomplished by a JUNCTION record
(which is a member of two one-to-many relationships).  In the Bachman
Diagram
, EMPOSITION is a junction between JOB and EMPLOYEE.  

  11) Bill-of-Materials – is a special relationship, where a record has
a "junction" record with two sets to it.  One set indicates a
parent option, the other set indicates a member option.  This allows
program to "explode or implode" the set relationship.  Two
examples: PARTS in a manufacturing company.  An airplane wing is a part,
but it consists of other parts, which in turn consist of other parts,
etc…  The second example is an org-chart, where Vice-President has 5
manager reporting to him, but he reports to the President.  Each manager
reports to a VP, but also has supervisor reporting to him, etc…

   12) Languages: Probably 90% of IDMS sites are COBOL, but IDMS
also supports PL1, Fortran, and Assembler, but not C.  See reference above
about IDMS’s own language called ADS.  IDMS also has a batch report-writer
called CULPRIT (which looks a little bit like RPG).

   13) OLQ = Online Query (which can actually also run in
batch).  You can use online menu-mode to make reports, save reports, or
generate a QFILE which is the code to reproduce the report.

   14) DMLO/DBOL – DMLO is CA’s took DBOL is a competitor tool from
Allen Systems Group. These allow online DML, retrieval or update.

   15) IDD – Integrated Data Dictionary – runs online or
batch.  Has its own language DDL (Data Definition Language) sometimes
called Double-"D" L (as compared to DDDL or Triple-"D" L
used for schemas).  The IDD itself is an IDMS database, sometimes called a
repository or metadata on other database systems.  Using DDL, IDD allows
you to store and/or view elements, records, users and limited access to
maps/programs.  Maps are usually stored using the MAPC compiler (online or
batch).  Some sites allow programmers to update the IDD, others sites do
not. 

    16) Program Error Codes – You can download my FREE IDMS
error code book.  (click here) IDMS Error Status codes are 4 characters, consisting of a
two byte major code and a two byte minor code.  The major code identifies
the verb type (Store, Obtain, etc…) and the minor code the error type. 
The most common errors are probably 0307 (end-of-set) and 0326 (record not
found).  You should test for these type of errors, and each program usually
has a catch-all routine for unexpected errors.  Programs can be written
using AUTOSTATUS or non-AUTOSTATUS.  AUTOSTATUS calls the error-checking
routine after each IDMS statement, so you must code the "ALLOWING"
phrase for any expected error messages.

   27) Database Navigation – using DML to navigate through a
database.  An Area Sweep ("Table space scan" for DB2 programmers)
is where you read the entire IDMS Database Area looking for occurences of
particular record types.  CALC-to-CALC sorted sets have a reputation of
being inefficient, since each member is on a different page, there a lot of
I/Os.  CALC-to-VIA is usually very efficient, usually only 1 or 2 I/Os,
because the data is clustered together. 

   28) Locking – is too detailed of a topic to go into here. 
Be aware that programs can lock each other out, slowing them down, or causing
deadlocks (error=0329).  There are shared (inquiry) locks and exclusive
(update) locks.

   29) What happens if you don’t COMMIT every few records in a
batch update program that updates hundreds of thousands of record?  Either
the lock tables will fill up (IDMS locks every record updated until that record
is committed) or the IDMS Journal will fill up.  See #30 below for more
info. 

   30) IDMS usually has 3 or 4 disk journal files.  These
record all the before/after images of all your updates.  Usually, IDMS
cycles through these files.  When one journal file fills up, an
auto-submitted job (via the WTO Write-To-Operator Exit) to offload the journal
to tape, and to condense the disk journal.  The utility can only condense
(remove) records that have been committed (see #29 above).    The
DBA can use the RollForward or RollBack Utility with the tape journal to restore
IDMS data.

   31) The IDMS log contains program dumps and statistical
information.  Remember that it is the JOURNAL, not the log, that contains
the before and after images used for recovery (see #30 above).  DB2 and
MS/SQL Server write before and after images to a file they call the LOG, but
IDMS calls the same type of file "The Journal", not the LOG.  Use a utility
called IDMSPLOG (IDMS Print Log) to print dumps or time ranges from the log
file). 




See also:
More Interview Questions

muzica noua