Actually the DASD themselves are easy to emulate when fully buffered and error free as IBM does today in its mainframes.
What is really hard to emulate is the Storage Control Unit and the appropriate set of CCWs associated with a specific Storage Control Unit and attached DASD. If I were doing this I might start with the 2314 DASF; it might run on any Selector or Block Multiplexor channel and since the tracks would be fully buffered in memory there really is no performance issue since there would be no seek time or latency time. For the command set to emulate see, Initial CKD implementation <https://en.wikipedia.org/wiki/Count_key_data#Initial_CKD_implementation> on Wikipedia or the 2314 reference manual. If there was more than one device on the channel u might have some serious hogging issues The 29MB/spindle or 232 MB/facility might be a limitation on some of the newer systems, but it is possible to modify IBM's Oss to change the characteristic of a DASD to almost anything by changing the device characteristics in a table of such - that's how the PDMs attached double capacity 2314s and 3350s; they redefined an unused device, e.g. 2311, to have the characteristics of these non-IBM supported devices. In this case all one might have to do is redefine the 2314 characteristics The full step which would be really really interesting would be to build an emulator that could operate as a Selector/2314/2318r, BMux/3380-2/3350 and/or (Bmux and Escon) /3990/3390. All together IBM implemented 83 CCWs for its CKD DASD, most but not all of which would have to executed by such an emuiator. Tom -----Original Message----- From: Paul Koning [mailto:[email protected]] Sent: Saturday, December 26, 2015 11:15 AM To: General Discussion: On-Topic and Off-Topic Posts Subject: Re: IBM channel-attached DASD emulation > On Dec 26, 2015, at 4:54 AM, Dave Wade < <mailto:[email protected]> [email protected]> wrote: > > ... > but IBM DASD are complex in that the physical block size can differ from track to track, and can be changed by the user. >From sector to sector, actually. And sectors can have key fields along with the data -- the device will find a sector by key match on a track. paul
