rbe@yrloc.ipsa.reuter.COM (Robert Bernecky) (02/08/91)
I think it was this group, or alt.folklore.computers, or whatever it's called, that someone asked why IBM 360 tape drives tie up the channel during rewind. This is not a 360 problem as much as an operating system access method coder problem. The 360 channel allows chaining of commands, so that you can, for example, write a disk command chain which will seek to a cylinder, search for a specific record on a track, and read/write one or more records from that point. If the commands were not chained, another process or cpu could move the arm between the commands, and you'd read gibberish. The device signals (channel end) when the data transfer associated with the operation has completed. Concurrently, or later, (device end) will be signalled indicating the device has completed any mechanical motion required, etc. For example, disk seek, tape rewind, and forms eject on a printer all signal channel end immediately, and device end much later. If command chaining occurs, it does not take place until device end time, and on old selector channels, ties up the channel during this time. The newer block multiplexor channels do not have this problem, as they can disconnect, under some circumstances. At any rate, what you observed was the rewind operation being combined via chaining with some other operation, such as unload, etc. If the rewind operation had been scheduled independently, the channel would have freed up during the rewind, and other tape operations could proceed concurrently with the rewind. Bob Bernecky Snake Island Research Inc. PS: I'm new to this posting biz, so please accept my apologies if I got the wrong group. I can't even FIND the name of the damn comp.filk group at the moment. Playing adventure is easier..
xxremak@csduts1.lerc.nasa.gov (David A. Remaklus) (02/14/91)
In article <1991Feb8.153055.26279@yrloc.ipsa.reuter.COM> rbe@yrloc.ipsa.reuter.COM (Robert Bernecky) writes: > (stuff deleted) > >At any rate, what you observed was the rewind operation being combined >via chaining with some other operation, such as unload, etc. If the >rewind operation had been scheduled independently, the channel would have >freed up during the rewind, and other tape operations could proceed >concurrently with the rewind. > Not necessarily. Dynamic disconnect and reconnect for channel protocol did not come until much later in the IBM 360/370 channel protocol. Under this protocol, certain devices (like disk and the new 3480 cartridge tape) will return a status byte and disconnect from the channel while the control unit and device service the operation (usually some sort of positioning operation). In the days of the older tape drives, the control unit, when it had to take part in the operation, was not able to disconnect from the channel so the channel stayed busy. The tape drive handles the rewind operation itself freeing the control unit. However, in the case of forward space and backspace type operations, the control unit takes part in identifying the tape marks and therefore does not return status until the operation completes. It is more likely that this is the type of operation that the individual had observed being performed. -- David A. Remaklus Currently at: NASA Lewis Research Center Amdahl Corporation MS 142-4 (216) 642-1044 Cleveland, Ohio 44135 (216) 433-5119 xxremak@csduts1.lerc.nasa.gov