Mike, You only think you turn off fast write.
I think the IBM 3990-6 and HDS 7690 were the last DASD controllers that really let you turn off DFW. The behavior you describe as "Waits for the record to be physically written before the next record is sent" in HDS parlance is a cache write through, which is usually due to a maintenance or error condition in a Mainframe DASD Array. Since the ICEBERG, HDS 7700C, IBM ESS 2105 and all EMC, DASD arrays have only ever supported write to cache with asynchronous destage to disk. EMC ignore the DFW status of a volume with DFW=OFF, whereas IBM and HDS put these writes in an earlier spot in the free queue. LSF by design would have written over one physical track, and then formatted a completely different track. My point is that no matter how you try to turn off cache, whether through setcache in IDCAMS or bypass cache in the CCW, it does not work. If you try to turn off DFW on a 2107 you get the error " IDC31562I THE DASDFASTWRITE PARAMETER IS NOT AVAILABLE FOR THE SPECIFIED SUBSYSTEM OR DEVICE ." Turning off DFW would have worked on DASD controllers fifteen years ago, but it does not work now. Ron > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:[email protected]] > On Behalf Of Mike Schwab > Sent: Thursday, December 13, 2012 12:25 PM > To: [email protected] > Subject: Re: [IBM-MAIN] dsf to write over entire volume > > Which is why we turn off fast write. Waits for the record to be physically > written before the next record is sent. > > On Thu, Dec 13, 2012 at 11:37 AM, Ron Hawkins > <[email protected]> wrote: > > Mike, > > > > Have you confirmed with each DASD vendor that the contents of the > > channel command generated by TRKFMT ERASDATA are actually passed to > > the disk drive for drive types, parity schemes and volume formats? > > > > To paraphrase your description, TRKFMT writes fulltrack records to > > cache then an empty track record to cache. The empty track record is > > all that makes it to the physical disk drives. > > > > In the case of Hitachi Dynamic Pools (HDP), if you erase a whole 38MB > > chunk like this there is nothing destaged and the chunk goes back in > > the free pool. > > > > Ron > > -- > 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: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
