For a key-sequenced data set, either specify an index control interval size or default to a system-calculated size. If you do not specify a size, VSAM calculates the CISIZE value based on the expected key-compression ratio of 3:1 and the number of data CIs per control area in the data set. After VSAM determines the number of CIs in a control area, it estimates whether the user-specified size is large enough for all the CIs in a control area. (See How VSAM Adjusts Control Interval Size.) If the size is too small, the system increases the size of the index control interval to VSAM's minimum acceptable size. If the specified size is larger than the minimum size that VSAM calculated, the system uses specified size.

You might need a larger CI than the size that VSAM calculated, depending on the allocation unit, the data CI size, the key length, and the key content as it affects compression. (It is rare to have the entire key represented in the index, because of key compression.) If the keys for the data set do not compress according to the estimated ratio (3:1), the index CI size that VSAM calculated might be too small, resulting in the inability to address CIs in one or more CAs. This results in allocated space that is unusable in the data set. After the first define (DEFINE), a catalog listing (LISTCAT) shows the number of control intervals in a control area and the key length of the data set.

   (KEYLEN/2) * DATA CI/CA ≤ INDEX CISIZE

The use of a 2:1 ratio rather than 3:1, which VSAM uses, allows for some of the additional overhead factors in the actual algorithm for determining the CI size.