Peter, I'm afraid you are being fed a bit of myth regarding random access and VSAM Compression with Extended Format. Compression/Decompression occurs at the record level, so you only decompress what you need. The myth is that you compress/decompress the whole CI just to touch one record - this is totally false.
Being a LZW based algorithm the cost of reading and writing is asymmetrical. Decompression costs much less CPU than compressing. Old measurements were around 6:1. Compression means you can use smaller buffer pools with BLSR and get the same hit rate - if that helps your current problem. I'm not sure I have figured out what you want to achieve, but there are some very good similarities to cache sizing on arrays. Your buffers may all be used, but some of them may have been used just once. Decreasing the size of the pool so that some CI are discarded by the LRU algorithm will not necessarily change the buffer hit rates. The general aim of buffer tuning is to achieve a target IO rate (not hit rate), which is how one would use VLBPAA. Whether you use BLSR or SMB, you may achieve your measurement aims by specifying the size of the pool in MB rather than number of buffers. If my memory is correct, the whole pool is getmained when the BP is first used, so it will be hard to track unused buffers from Virtual Storage statistics. BTW, are you using RMODE31 parm to put the buffers above the line, and DEFERW if you are updating the file? Ron > > 1. I have been told that REPRO is a "good" usage of compressed data -- > repeated sequential access of any kind gives good results because each > buffer is uncompressed only once. OTOH, nearly random access and > compressed > are "not good" use of compressed data (i.e., higher CPU utilization due to > repeated decompression). My particular application is in the "nearly > random" category. > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
