[EMAIL PROTECTED] wrote:
> Would the distinction between "warm start" and "cold start" for OS/370
> then be whether these queue files are re-initialized?  I remember
> both terms being used with these systems, but not what the distinction
> between them was.

for cp67 & vm370 the "spool file system" .... basically emulated unit
record information was subject to cold and warm start ... and then
chkpt start was later added.

part of booting the system including bringing up the spool file system
... and  was prerequisite to the system coming up .... so it had to
come up. in part, the "spool file" disk area tended to be shared
between spool files and dynamic virtual memory paging. if the area
wasn't initialized and available ... you also didn't have paging ...
integral to the operation of the system.

oiriginally a cold start was that it initialized the area to completely
empty (all existing data was lost). warm shtudown would write
in-storage allocation information to disk at shutdown  typicalland on
boot, this information would be reread back into memory. on things like
power-failure ... the system went down w/o having written the warm
start information and had to come up cold. cp67 also added automatic
creation of a "dump file" (in the system spool area) on a kernel crash
... and then automatic reboot (kernel failure, dump, save warm data,
reboot, and warm start could be done in a minute or two.

there is this story about somebody at mit making a kernel change to one
of their systems in an i/o driver and having 26 system crashes and
reboots in the course of the day. the story is that this prompted some
improvements in multics ... since multics at the time supposedly might
take an hr or so to come back up after a crash
http://www.multicians.org/thvv/360-67.html
.
part of this was that the cp67 & vm370 stuff was done at the science
center on 4th flr of 545 tech sq
http://www.garlic.com/~lynn/subtopic.html#545tech

and multics was being done on the 5th floor.

to handle the power failure case, chkpt start was added. during normal
operation, a very abbreviated vesion of some in-storage control
information was written to disk. if reboot selected chkpt start ...
boot would read the abbreviated information and then use that to reread
the spool area to recreate the rest of the information.  this was
somewhat analogous to fsck since the reread/reconstruct could take an
hr for a spool system with a large number of files.

the automatic reboot would bring the system back to the point where
users could log back in. however, over the years, lots of automated
processes were added to normal operating environment ... and these
required manual initialization.

for the automated benchmarking process, mention in recent post
http://www.garlic.com/~lynn/2005s.html#44 winscape?
http://www.garlic.com/~lynn/2005s.html#45 winscape?

i had added a boot process that could be used to start up these
automated processes at boot as part of normal production operation. the
purpose for automated benchmark
http://www.garlic.com/~lynn/subtopic.html#bench

was to allow rebooting between the automated benchmarking ... to have a
clean, fresh startup between each benchmark. the automated benchmarking
could run for days unattended ... with automated reboot and starting
each new benchmark.

this feature was incorporated in standard released system and is now
part of standard production vm operation.

later doing some high-speed networking support as part of hsdt
http://www.garlic.com/~lynn/subnetwork.html#hsdt

i did "spool file system" rewrite ... recent lengthy post discussing
characteristics of sfs rewrite
http://www.garlic.com/~lynn/2005s.html#28 MVCIN instruction

it was primarily to significantly improve the performance ...  but
while i was at it ... i changed the paradigm so that spool areas could
be added & removed during normal operation (w/o rebooting the system).
this had the side-effect that even if chkpt-like start was required
.... the system reboot didn't have to be delayed until it finished (i
also redid how the chkpt operation was implemented so that it was
significantly quicker).

Reply via email to