All topics
43 / 60
VSAMbeginner

VSAM KSDS: Key-Sequenced Data Sets

Understand the structure and use of KSDS, the most common VSAM organization, and how its index and data components work together.

A KSDS stores records in key order and can find any one of them directly by key. It is what sits under almost every online file on a mainframe.

A KSDS is a library that keeps books shelved in call-number order (the data component) alongside a card catalog (the index component) mapping each call number straight to its shelf location — you can browse the shelves start to finish, or walk directly to exactly the book you want without touching any other shelf.

Key Concepts

1
Two components, and both exist on disk.
    index component   maps a key to a location in the data component
    data component    the records themselves, in key order
2
The index is a tree. Finding customer C0004521 in a file of ten million reads a handful of pages, not ten million records.
C0004521
3
It supports both access patterns, from one open file.
    direct       jump straight to a key
    sequential   read in ascending key order
    dynamic      both, in the same program
4
That combination is what makes it the default choice. A batch job reads the whole file in key order; an online transaction fetches one record; they use the same file.
5
Free space is the parameter people get wrong. A KSDS keeps records in key order, so inserting a record means putting it in its correct place — not at the end.
6
    FREESPACE(20 10)
              |  |
              |  percent of each control area left free
              percent of each control interval left free
7
With free space available, an insert fits into the gap in the right control interval and costs almost nothing.
8
Without it, you get a split. The control interval is full, so VSAM allocates a new one, moves half the records into it, and updates the index.
9
    before:  [ A B C D E ]  full
    insert F
    after:   [ A B C ] [ D E F ]   two CIs, index updated
10
One split is cheap. A file loaded with no free space and then heavily inserted into splits constantly, the records end up physically scattered despite being logically in key order, and sequential reading — the thing key order was supposed to make fast — turns into random I/O.
11
A control area split is the same thing one level up, and is considerably more expensive.
12
Which is why a heavily inserted KSDS is periodically reorganised: unload it, redefine it, reload it in key order with free space restored. That is the VSAM equivalent of a DB2 REORG, and it exists for the same reason.
REORG
13
The trade against a flat file. The index costs storage, and every insert, update and delete must maintain it. What it buys is finding one record among millions without reading them all — which a sequential file cannot do at any size.
14
What the interviewer is probing.1. "What makes a KSDS different from the other VSAM types?" Probing: the fundamentals. Stalls: "It has keys." Moves up: records are stored and retrievable by a unique key in sequence, through an index component separate from the data component.
15
2. "What is a control interval split?" Probing: the mechanism. Stalls: "A problem." *Moves up:* when an insert does not fit, VSAM splits the control interval, which costs time and leaves the file physically disordered.
16
3. "How do you reduce splits?" Probing: the tuning. Stalls: "You cannot." Moves up: define free space with the FREESPACE parameter so inserts fit, and REPRO or reorganise periodically.
17
4. "Can the key be changed on an existing record?" Probing: the detail. Stalls: "Yes." Moves up: no — a key change is a delete and an insert, because the key determines placement.
Use cases:ksdsvsam-basicsindexed-filesdirect-access