munck@Mitre-Bedford (02/25/85)
An important indicator of the nature of the IX/370 to VM relationship
is the support by VM of a generic 4096-byte random access device. This
means that IX/370 code can assume that all of its disk storage is on a
single kind of device, that the device is extremely simple to drive, and
that it gets only a few kinds of errors, for which no error recovery is
possible. Therefore ALL of the UNIX kernel (and elsewhere code) that
deals with multiple kinds of disks and that tries to do error recovery
or otherwise handle "soft" errors CAN BE ELIMINATED. (Someone implied
that this is a recent feature of VM; we had it in our CP/67 system at
Brown Univ. about 12 years ago.)
VM provides an "ideal" virtual machine to guest operating systems in
this and a number of other ways. Remember, too, that IBM operating
systems are REALLY GOOD AT I/O. For all of our complaints about the
rottenness of JCL and the horrors of TSO, those IBM mainframes and OSs
with their dedicated block channels, nth-level interrupt handlers, and
seventy-eleven kinds of mass storage can really move the bits in and
out. Anyone have a UNIX that's I/O-bound? IX/370 just might be
substantially better than whatever you've got now.
-- Bob Munck, MITRE