Depends on the design/implementation of the O/S. It certainly used to make a big difference on VM in the early days when the Master Processor could be a major bottleneck at times. The constraints were removed over time and the last were finally lifted in V5, which was where the 2GB line constraints were also lifted. There used to be other dependencies on the M/P than I/O. Services like storage management come to mind. Anything that had to be single-threaded was fair game for Master Processor work.
In the early days of multi-processor configurations of TPF, it was possible for any given processor to be devoted to I/O. IIRC, our TPF Systems people (at Piedmont Airlines) made CPU 0 the base supervisor, including all supervisor functions except for I/O, CPU 1 was devoted to I/O and other CPUs devoted to applications. In this case, CPU 1 was not completely devoted to I/O, applications were allowed to run there when there was time available; however, I/O took precedence and would preempt application work. I don't know about the design of z/OS. I hope that there aren't too many dependencies on the Base Processor. I'll ask our z/OS sysprogs. Regards, Richard Schuh > -----Original Message----- > From: The IBM z/VM Operating System > [mailto:[email protected]] On Behalf Of Don W. > Sent: Wednesday, December 09, 2009 5:21 AM > To: [email protected] > Subject: Base Processor? > > When you set up the directory entry for a z/OS guest with > multiple virtua= l CPU's (e.g. CPU 00, CPU 01) and you QUERY > CPUS, CPU 00 shows as BASE. Wha= t is the difference between > a BASE CPU and a non-BASE CPU? Does it really m= ake any > difference in this environment. In terms of real processors, > at one t= ime the base processor had to do the I/O. I do not > believe that is true any more. Am I correct? and when did > that stop being the case? >
