Various kinds of System reloads 2361
No. A SYSGEN is never required to restart the system in the same configuration it had the last time it was running. But on a large OS-370 system with remote job entry (JES2 or JES3), TSO (Time Sharing Option), CICS (transaction processing monitor) etc, a reboot (IPL) does take anywhere from 5 to 30 minutes, as each layer of the system is brought up, asks questions to decide which queue files must be recovered and which should be re-initialized ("Format Spool Cylinders yes or no?") and eventually declares itself ready to provide services. This initialization is allowed to proceed one component at a time in order to ensure that disk space and main storage does not get unnecessarily fragmented, thus reducing performance once the system is completely up. (My hands-on experience with this was on an S360-75 with OS-360-MVT release 10 and HASP-II. Things may have gotten more relaxed and thus faster once the OS got virtual memory, so that core fragmentation was no longer an issue. Do people even bother to tune the link pack area any more?)
for cp67 & vm370 the "spool file system" .... basically emulated unit record information was subject to cold...
I think this is a little imprecise. In terms of instruction types, x86 has been usably extended to 64 bits...
Yes, I do know that well, having spent most of my DEC-related years at companies that supplied add-on hardware for which I those drivers to be "site-specific system patches".
For *n*x; not really. It now seems the x86 architecture has run out of steam in terms of speed. The field of compebreastors has been steadily narrowed...
devices. At boot time, system initialization would call each device driver at its "probe" entry point, and the driver would look to see if the device that it supported did in fact exist. If it did, it would build the data structures for the device and link them into the relevant system chain; if not, it would return an "offline" indication.
The DEC-supplied "starter system" supplied on the distribution media would typically contain a configuration with "one of each" of the major tape and disk devices. This allowed you to get the software off the distribution media onto a disk drive of whatever type you had on your site, and then perform a SYSGEN to get rid of the devices you did not have and add the additional ones that you did that were not in the starter system.
Unix had the same mechanism.
Device drivers could be linked into the kernel image, or they could be loadable.
As memory prices went down, there was less incentive to spend time tweaking the system to get rid of drivers not needed, and more incentive to keep all installations as similar as possible.
Not having spent time in a TOPS-10 shop, I will cede to your judgement on that system, but my experience was that Unix systems were much easier to manage than the IBM mainframe systems I saw. For example, IBM allowed you to tune the BLKSIZE on each file on a disk, requiring users to know what optimal values would be for each of their files, while Unix only had one file type: A byte stream formatted onto 512-byte disk sectors, so that all file storage was interchangeable.
Lars Poulsen
Various kinds of System reloads 2362
Alt Folklore Computers from Newsgroups