Various kinds of System reloads 2355

My recollection (of installing versions of MVS about 20 years ago) was that - as Peter says - sysgen was an early stage in building a new version of the OS.

Since the stage 2 of sysgen was an enormous jobstream, it typically ran overnight. It absolutely was not part of a boot (IPL) process. A sysgen was done about once every 9-12 months at the very most, but - in those days - IPLs were done several times a week.

When we built an initial MVS system this way, the first IPL typically came a few days later - for a start we had to check out all the warning messages that were produced by the gen. For example, there'd be linkedit warnings for bits of the OS that we'd chosen to miss out, but of course there'd also be warnings for things we accidentally left out. There was no point in going ahead with all the next part of rebuilding a complete system if we'd got the initial bit wrong...

Once we were happy with that and had done a few other things, we'd have the first ever IPL, on a test machine, of the new OS. If it worked it'd be used as the basis for several months worth of customisation, installation of vendor applications, customisation of those, and lots and lots of system, application, user & disaster recovery testing.

Only if we got through all of that would the system be put into real use, and even then it started off supporting development (eg TSO) users, and a limited amount of production batch work. Only after that lot seemed ok (perhaps after a month, say), would we consider starting the production workload on it.

So sysgen was not part of the boot process. Maybe you mean something else, eg NIP - Nucleus Initialisation Programme - which basically loaded the dyanmically changeable parts of the nucleus (eg device configuration, created by doing an IO GEN) alongside the static parts (ie what had been built by sysgen)?

-- Jeremy C B Nicoll - my opinions are my own.