A CI split (unless it forces a CA split) involves very little extra overhead - like typically an extra Data CI write and Index CI write. A CA split on the other hand means you have to physically move half the Data CIs in the old CA to another CA. For a 3390 cylinder CA and 4KiB Data CISIZE, that means an extra 90 blocks read into and 90 blocks written out of your application buffer pools, which can still affect on-line response if this happens too often.
Prior to DASD subsystems with large amounts of cache, you could also pay a read-performance hit when reading areas that had undergone CI or CA splits, but modern caching subsystems do a good job of hiding that penalty. Everything depends on how the dataset will be used and the pattern of update activity to it. I would be inclined to suspect there are few cases today where CI-freespace makes sense, but we still have cases where we specify CA-freespace because it seems to help. JC Ewing On 11/10/2009 06:55 AM, Barry Jones wrote: > Hello, > > In the old days,CA splits were quite detrimental to performance of VSAM > applications. > > In today's world with lots of cache in modern DASD systems, are CA splits > still a concern? > Or should I just allocate KSDSs with FREESPC(00 00) and save the space? > > All thought appreciated. > > Barry. ... -- Joel C. Ewing, Fort Smith, AR [email protected] ---------------------------------------------------------------------- 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
