Explain DD Statements and Dataset Allocation in IBM Mainframes

13 / 60

JCLbeginner

DD Statements and Dataset Allocation

Understand how DD statements connect a program's logical file references to physical datasets, devices, and allocation parameters.

A DD statement connects a name the program uses to an actual dataset. It is the most-tested thing in JCL because every job question eventually reaches it.

A DD statement is like the shipping label and handling instructions taped to a box — it doesn't just name the box (DSN), it tells the courier whether to open a brand-new box, reuse one already in the warehouse, and what to do with it if the delivery truck crashes en route.

Key Concepts

    //CUSTFILE DD DSN=PROD.CUSTOMER.MASTER,DISP=SHR

CUSTFILE is the DD name — what the COBOL ASSIGN clause or the utility's control statement refers to. Everything after it says which dataset and under what terms.

CUSTFILEASSIGN

Every DD answers three questions.

    DSN=    which dataset
    DISP=   what state it is in, and what to do with it afterwards
    SPACE=  how much room, if it is new

A new dataset needs more.

    //OUTFILE DD DSN=PROD.DAILY.EXTRACT,
    //            DISP=(NEW,CATLG,DELETE),
    //            SPACE=(CYL,(50,10),RLSE),
    //            DCB=(RECFM=FB,LRECL=200,BLKSIZE=0)

SPACE=(CYL,(50,10),RLSE) is 50 cylinders to start, 10 more each time it fills, and release what was not used. Under-allocate the secondary and the job abends part-written with a space abend — one of the most common batch failures.

SPACE=(CYL,(50,10),RLSE)

DISP is where most mistakes happen, and it has three parts:

DISP

    DISP=(NEW,CATLG,DELETE)
          |   |     |
          |   |     what to do if the step ABENDS
          |   what to do if the step ends NORMALLY
          current state

Read that as: create it; if we finish, catalog it; if we crash, delete it. That last part is the point — it stops a half-written file being left where a later job might read it as complete.

Other things a DD can be.

    //SYSPRINT DD SYSOUT=*           print output, viewed in SDSF
    //SYSIN    DD *                  instream data follows
      control statements here
    /*
    //SYSOUT   DD DUMMY              discard it entirely

DUMMY satisfies a program that insists on a file while doing no I/O at all — useful for suppressing output you do not want.

DUMMY

Concatenation: several datasets read as one. Omit the DD name on the following statements:

    //SYSLIB   DD DSN=MY.COPYLIB,DISP=SHR
    //         DD DSN=PROD.COPYLIB,DISP=SHR
    //         DD DSN=SYS1.MACLIB,DISP=SHR

They are searched in order, which is exactly how a local copybook overrides the production one — and exactly how a stale library in the wrong position silently wins.

The operational point worth making. DISP is a contract covering both the success and the failure path. Writing DISP=(NEW,CATLG,CATLG) means a crashed, partly-written dataset is catalogued and left behind for somebody to find later and trust.

DISPDISP=(NEW,CATLG,CATLG)

What the interviewer is probing.1. "What does a DD statement connect?" Probing: the fundamentals. Stalls: "A file." *Moves up:* the program-internal file name to a real dataset, which is why the same program can run against test and production data unchanged.

2. "What do SPACE parameters mean?" Probing: the detail. Stalls: "The size." Moves up: primary and secondary allocation in the given unit, plus a directory block count for partitioned datasets — too small a primary means an abend on extend.

3. "What does DISP=(NEW,CATLG,DELETE) say?" Probing: practical reading. Stalls: "Create it." Moves up: create it now; catalog it if the step ends normally; delete it if the step abends — the third value is why failed runs leave no half-written file.

4. "What is DD DUMMY for?" Probing: the technique. Stalls: "Nothing." Moves up: satisfying a DD the program opens but you do not want written — reads hit end of file immediately and writes are discarded.