On Thu, 6 Oct 2005, McKown, John wrote: > Well, from my viewpoint, there is one very big minus to doing the ECKD > emulation on the host (mainframe). How much CPU will it cost me? ...
Yes! yes. I acknowledge David's points in rebuttal to my comment. But YOUR point is a big part of why I don't like host-side ECKD. ECKD emulation is a double-whammy: making fixed-block storage look like count-key, and then flattening out the count-key into fixed blocks (for CMS, Linux, even VSAM). So not only are there two places where CPU is consumed, but there is also wasted storage. But walk it all the way through. Perhaps you'll agree. In any case, you'll be armed. In the balance, [E]CKD has been an artificial requirement for almost two decades. I remember when 3390s came out (REAL 3390s; remember?) that some IBMers said they would be the last of that kind of disk. Could not imagine then what they were talking about, but now I see: physically trackable disk was going away. Fixed block is here to stay, and apps which need ECKD function now get it via emulation ... ALWAYS. I've been saying for a long time that the storage vendors should give us fixed-block on the channel. But I'm just an old academic and unskilled with respect to the "business case". But now, the business case has arrived without my help: SAN. So now, SAN is fixed block by nature, and those few apps and systems needing ECKD are forced to either not play or to do their own ECKD emulation. The day will come with z/OS will bear the burden for its own ECKD emulation. That will incent customers and IBMers alike to avoid CKD gymnastics where they're not really worth it. Dave Boyes suggests that z/VM have the same capability. And why not? He's right in so far as VM is the migration aide for all things Z. Methinks that Endicott could lift code of Poughkeepsie conception and drop it into the same EDEV arena: emulated ECKD with z/VM. If that's all it takes, I say, "Go for it!". ... > Can this processing be done in some other "special purpose" CP > (like an IFL or zAAP) so that my OEM software costs are not impacted > when I need more "dasd emulation" power? ... Here's a thought: let z/VM use zAAP in the hypervisor for EDEV! (And John McKown gets credit for the idea.) -- R; ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
