On Thu, Mar 24, 2011 at 12:37 PM, Ron Hawkins <[email protected]> wrote: > Shmuel, > > Blockset may resemble paging IO in the way it would track switch over a CYL > as it laid down blocks, but that's probably not an advantage any more with > cache all storage and CKD encapsulation into FBA. This behavior is one > reason why the cache hit rate for Local Page datasets is somewhat poor, but > I haven't observed how block paging behaves in polluted or small cache. > > DFSORT's IO to the SORTWK are explicitly requested in response to DFSORT > deciding it has exhausted its own memory. At startup it attempts to, and > usually succeeds in selecting memory boundaries that avoid paging, using the > SORTWK as the memory overflow area. This could be viewed as DFSORT doing its > own paging, or at least its own memory management. > > Otherwise AFAIK DFSORT will have pages stolen and demand page with or > without block paging just like any other batch job. > > I'm sure you know all this already, so what's your point? > > Ron
The DB2 people wanting 150GB in paging packs refused to code SORTWK DD statements to force in-core sorts. I have always coded SORTWKs even if they were never needed. -- Mike A Schwab, Springfield IL USA Where do Forest Rangers go to get away from it all? ---------------------------------------------------------------------- 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
