Pinnacle wrote: > > Tom, > > I tested this in the old days of SLED DASD, and the directory was a > keyed track and you COULD NOT store a member in the directory track. It > always took 2 tracks minimum for a PDS (of course, this has not been > true for about 15 years). > > Regards, > Tom Conley > ----------------------------------------------------------------------
Tom, I agree with (the other) Tom - I don't think this has ever been true. I just tried making a member in a new single track PDS under MVS 3.8 and it worked without a problem - and that software is more than 20 years old. There would be an increased statistical likelihood that the first member would start at the start of a track because it is possible that there is not enough room after the directory to fit a data block on the last directory track. But there seems to be nothing to prevent it in general - which is good because that matches my recollection from those times. BTW, on the cylinders vs tracks allocation point, I think that if a data set is allocated in cylinders then the access method can use the MT bit (the Multi Track bit is the x'80' bit in the CCW opcode) when building search channel programs, thus reducing CPU overhead and the chance of RPS miss. For example, for search-key-equal-or-high, a x'69' CCW can search a track, while a x'E9' CCW can search till the end of the cylinder. Note that the end of extent being on a cylinder boundary is important, not that the extent starts on a cylinder boundary. At least that what I've heard. Mind you, I think this is all superceded by ECKD now-a-days. Cheers, Greg ---------------------------------------------------------------------- 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
