Beginners often master control intervals first because CISZ appears explicitly on DEFINE CLUSTER. Control areas are the next layer up: each CA is a contiguous range of disk space that VSAM treats as a single administrative region containing a whole number of CIs. CA size therefore answers the question, "How many CIs move together when VSAM does certain kinds of work?" It is tied to geometry—tracks and cylinders—and to CISZ, because the product of how many CIs fit on a track and how many tracks participate in a CA determines how wide each CA is in terms of intervals. You rarely type a parameter called CASZ in basic JCL the way you type CISZ; instead, CA size emerges from allocation choices, device type, and CI size. Understanding that emergence is what lets you reason about CA splits, locality, and capacity without memorizing every internal formula. Storage administrators still speak about CA size when they tune batch versus online files because a CA split touches every CI that VSAM decides to rebalance across an old CA and a new CA, and the number of CIs in the CA scales that cost.
When you allocate CYLINDERS or TRACKS, the system gives VSAM raw capacity. VSAM divides that capacity into CIs of length CISZ, subject to alignment rules that depend on the device. Those CIs are grouped into CAs. On many teaching examples, one CA corresponds to one cylinder for 3390-class devices, which makes mental models easier: each cylinder holds N CIs where N is roughly the bytes per cylinder divided by CISZ, rounded according to VSAM rules. If you change CISZ, you change N even when the cylinder count stays the same, which changes how many CI splits can occur before you exhaust the free CIs that FREESPACE reserved in that CA. That interaction is why CA size discussions are inseparable from CISZ and FREESPACE tuning rather than from primary allocation alone.
| Term | Meaning |
|---|---|
| CA boundary | The end of one control area and the start of the next. Inserts that walk in key order may cross CA boundaries as the file grows. |
| CIs per CA | Total bytes in the CA divided by CISZ (conceptually). Larger CISZ means fewer CIs per CA for the same CA byte length. |
| FREESPACE CA% | Percentage of CIs in each CA left unused at initial load so later CI splits can use free CIs without a CA split. |
A CA split happens when VSAM needs a CI split but there is no empty CI in the current CA. The system allocates another CA, moves about half of the CIs from the crowded CA to the new CA, updates the index, and then completes the CI split. If each CA contains many CIs because CISZ is small relative to track size, a CA split touches many intervals and can run for noticeable time under heavy insert load. If each CA contains fewer CIs, the split still reorganizes keys but moves fewer intervals per event—yet you may hit CA boundaries more often depending on growth patterns. You do not pick CA split behavior directly; you pick CISZ, FREESPACE, and allocation, then observe split behavior. EXAMINE and performance reports help distinguish CI split storms from CA split spikes.
FREESPACE takes two percentages in the common teaching form FREESPACE(ci-percent ca-percent). The first reserves free bytes inside each CI at load time to delay CI splits. The second reserves whole free CIs scattered through each CA so that when a CI does need to split, VSAM can allocate a new CI inside the same CA instead of escalating to a CA split. If you set the CA percentage to zero, you are declaring that you are willing to accept CA splits as soon as CI splits need another CI and none are free in that CA. Many growing KSDS files use a non-zero CA percentage even when the CI percentage is modest, specifically to buy insurance against CA splits during business hours.
12345678910111213DEFINE CLUSTER ( - NAME(USERID.CUSTOMER.KSDS) - INDEXED - RECORDSIZE(500 500) - KEYS(16 0) - CISZ(8192) - FREESPACE(10 15)) - DATA (NAME(USERID.CUSTOMER.KSDS.DATA) - CYLINDERS(50 20) - VOLUMES(SYSDA)) - INDEX (NAME(USERID.CUSTOMER.KSDS.INDEX) - CYLINDERS(2 1) - VOLUMES(SYSDA))
In this sketch, CISZ 8192 controls CI density, while FREESPACE(10 15) combines intra-CI headroom with inter-CI headroom at the CA level. Changing only CISZ changes how many CIs fit in each CA and therefore how quickly the CA-level free CIs are consumed even when the FREESPACE percentages stay the same, because percentages apply to different underlying counts.
If a CI is a small box, a CA is the shelf that holds many boxes. The shelf size depends on how big each box is and how much shelf space you bought. If every box on a shelf is full and you need a new empty box, you might need a whole new shelf—that is a CA split. Leaving a few empty boxes on each shelf (FREESPACE CA) means you usually fix problems without buying a new shelf.
1. A CA is made up of what smaller VSAM units?
2. Which parameter helps reduce CA splits by reserving unused CIs in each CA?
3. If CISZ increases but CA byte length stays similar, what happens to CIs per CA?