On 8/24/2013 10:06 AM, John Gilmore wrote:
A preoccupation with having a first PDS member begin on a notional track boundary of an emulated device strikes me as misplaced, even bizarre; but it would be my guess that it is also benign.
I agree, but only when it comes to general considerations (see below).
To take it seriously for the moment, It is achievable by calculating the number D of directory blocks that fit on an emulated 3390 track and then always allocating some integral multiple of D directory blocks. (This calculation is trivial, but I have suppressed its result here to avoid encouraging its misuse.)
This is far from trivial. I just ran some tests, and an allocation of the number of directory blocks you posit, on a 3390, will place the directory EOF block on the next track. There were some device types where there was enough space on a track to place the EOF, but on the 3390 there isn't, and if you change the allocation to one block less, there is not only room for the EOF, but possibly also a (small) block of the first member.
I do not myself think initial-member track alignment should be sought. What was perhaps optimal for a spinning physical 3390 DASD will only adventitiously be optimal or even appropriate for an emulated one.
Modern controllers I'm familiar with operate on a track basis, whatever the hardware or firmware does internally. It is therefore plausible than I/O for data that will fit on a single track will be faster than if the same data were split between tracks. There may be extremely critical (e.g., real-time) systems for which such placement considerations are anything but bizarre.
Gerhard Postpischil Bradford, Vermont ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
