Hi all,

I'm currently studying various kinds of compression in several databases, and I kinda confused in index compression at DB2 OS/390 version 9.5, it supposed to use a "key map" to replace dictionary, right ? But actually what happened in the process ?

In my imagination, if there is 4 similar rows, 3 of them would be eradicated and the last one will be the key to expand the other 3 when needed, so, what does the key map do ???

Are you sure INDEX can be COMPRESSED in V9.5 in OS/390? In LUW, it's not possible.

I searched IBM Redbooks
http://www.redbooks.ibm.com/
with keywords '"DB2 9" AND z/OS', then I got some interesting documents.

14 results in Redbooks, Redpapers, Drafts and Technotes
http://www.redbooks.ibm.com/cgi-bin/searchsite.cgi?query="DB2+9"+AND+z/OS

, especcialy I was interested in
Index Compression with DB2 9 for z/OS
http://www.redbooks.ibm.com/abstract...4345.html?Open

Yes, it can. Actually, compression is only applied to the leaf pages, and it uses a fixed algorithm, i.e., one that does not depend on a compression dictionary. (Or stated otherwise, it depends on a single, fixed, built-in dictionary.)
After compression, the leaf pages must have been reduced to (less than) 4K, which was the (uncompressed) index page size limit in older versions of DB2 for z/OS. Through compression, the actual, uncompressed size of the leaf pages can now be 8K, 16K or 32K.

Thanks Peter. Never worked on DB2 Z/OS, but great to know that stuff about Index-Compression.