Keyboard Shortcuts
ctrl + shift + ? :
Show all keyboard shortcuts
ctrl + g :
Navigate to a group
ctrl + shift + f :
Find
ctrl + / :
Quick actions
esc to dismiss
Likes
Search
Issue with HRC075dk (Alternate nucleus)
|
Why should we need to change DMKSYS depending on where we want the nucleus loaded and IPL'd??
?
The whole idea of "alternate nucleus" is to be able to test a new prod nucleus and be able to fall back if it fails.
?
Another important concept is to be able to place exactly the same code in production that was previously tested WITHOUT ANY ALTERATIONS.
?
IF you look at the code updates suggested in Melinda Varian's "What Mother Never Told You..." document, you will see that the idea is to simply separate the IPL volume as being different than the SYSRES volume... the nuc load deck would still be used to load onto the same cuu/cyl as specified in DMKSYS
?
The latest version of "alt nuc" that I implemented in 199x? (HPO, SEPP??) also made the cyl location definable by volid.
?
I think it not ok to have primary and alt nucs needing separate dmksys content
?
cheers
William |
|
William,
?
It seems to me that you are misunderstanding how to use the alternate nucleus arrangement.
?
First off, DMKSYS is a table of parameters; the resulting assembly doesn't contain any code.
?
If you want to test an alternate nucleus without any changes to a production system, you can do just that.? You still have to change DMKSYS to point to a new SYSVOL and new SYSRES (so that the system loader knows where to write the new nucleus), but your original production system nucleus would be unchanged as long as you specified a different SYSVOL & SYSRES.
?
If you needed to go back to a stable nucleus, you simply IPL from your previous sysres.
?
If you want to set it up like Melinda Varian's example, you can still do so.? But regardless, you still have to alter DMKSYS to point to the SYSVOL and SYSRES that you want the loader to write the nucleus to.
?
Note that the alternate nucleus is simply that: another copy of a CP nucleus on another volume.? It is not a copy of another set of checkpoint, error recording, and warm start areas.? If you want those to be unique on their own volume, you can set that up as well, but it really isn't part of the alternate nucleus support.
?
Finally, in VM/SP back in the day, you were allowed to specify SYSVOL=*, so that you didnt have to key in the alternate's volser, but you still had to change the SYSRES= value to the alternate dasd CUU hosting the alternate nucleus location.? SYSVOL=* is not allowed by HRC075DK.
?
Regards,
Bob
? |
|
On Thu, Aug 15, 2024 at 02:55 AM, William Denton wrote:
Why should we need to change DMKSYS depending on where we want the nucleus loaded and IPL'd??It is important to remember, we are playing with an operating system from the late 1970s through early 1980s, about 40 years ago, which IBM got no money for developing. It only aided in hardware sales. As such where would your emphasis be? ?
As I have time, I read through Lynn Wheeler's (creator of VM's Wheeler Scheduler) e-mails and notes and it's amazing to me we got a system such as we did.
?
?... Mark S.
? |
|
Mark...
?
Yes, Lynn was one of the greats among many... we collaborated on the scheduler changes that ended up in SEPP based on some the observations my shop was seeing trying to run a large interactive guest terminal application in a virtual machine. The requirement to immediately dispatch a relatively localized working set into play based on an interrupt from an UNSP 2703 line was a challenge.
?
One of my original attempts as improving things was the MINRES mod that was originally intended to make sure that the latest working set for a VM was resident at all times. It was later used as a way to restrict a VM to a certain number of resident pages... not my original goal... John Alford made this alternative popular.
?
We were also working with Dave Newsome and Peter (?? can't remember his last name from Cambridge Sci Ctr) on these efforts.
?
So, finally, I am in the near-testing stage of my version of Alternate Nucleus which is based on:
* Separating the concept of the IPL(nucleus) volume from the SYSRES(other stuff) volume
* Not requiring nucleui to be located on the same device types or at the same cylinder location
* Being able to load the exactly same load deck to alternate locations without reassemble or rebuild
?
I am not looking to replace HRC075DK which does not actually have any impact on nucleus processing. This can still be used to place WRM/ERR/CKP on any other volume
?
cheers,
William Denton
?
ps.. how do you happen to have access to Lynn's email & notes.. just curious |
|
Hi, My first job out of college was with Texas Instruments.? I chose TI for two reasons: 1)? It was my understanding the BSEPP and SEPP work were done at TI main site. 2)? The job was to be the first and lead systems programmer at the Lewisville Interactive Computer Center (LICC).? The logo was the Rolling Stone Tongue logo. After getting it all put together, getting an exceeds expectations review someone moved me to another group.? I worked on a SEL 32/55 for six months or so and got another exceeds review.? It was at this point I was told they were moving me to work on the Ada compiler. All I wanted to do was learn VM from those who wrote BSEPP and SEPP. A friend from college recommended me for with Owens Corning Fiberglas.? I took it not for the 60% raise, which was really nice.? I took it because I was again working with VM and as a lead systems programmer with a team. We supported more than VM, but it was my baby. To read about BSEPP and SEPP is great!? I wanted to work beside and learn from those great people.? Never getting to even meet them is one of my many regrets. Bertram Moshier WB8ERT On Sat, Aug 17, 2024, 02:16 William Denton via <williamedenton=[email protected]> wrote:
|
|
On Sat, Aug 17, 2024 at 02:16 AM, William Denton wrote:
ps.. how do you happen to have access to Lynn's email & notes.. just curiousSee Bob's response. I never talked to him, but heard much about him at the few SHAREs I was able to attend. Ran into his e-mails on a discussion list and he referenced the web site. ?
AMAZING!
?
?... Mark S. |
|
Yes, of course.. Dick... senior moment there..?
?
Those heady days of the mid-70's thru the 80's were great times for VM. I was the lead VM systems programmer at Home Savings at the time. We were running an $8B bank on a 3mb 158 under VM with DOS r25 guests along with the normal 50-80 CMS users. We had an online network of around 120-ish teller terminals. (The 158 got upgraded to a 168 and then a 3083 and onward.)
?
The performance challenges of that online terminal network was one of the main motivations for us to engage with Lynn and others to discuss the rumored "Wheeler Scheduler" (which never made it into a shipped version of CP67) and how its concepts might address those challenges. These discussions along with other Share members resulted in the changes to VM that ended up with the features included in SEPP and, later, HPO
?
Great times.. nice to think back on all we accomplished in those days.
?
Cheers,
William Denton
?
ps.. for the younger people here, you need to find a copy of the LSRAD report that "saved" VM from the MVS goliath |
Mark A. Stevens
to navigate to use esc to dismiss