• Please review our updated Terms and Rules here

Help Please: IBM 370 Mainframe System Operator Console

R

robcarnegie

Member
Joined
Jun 5, 2016
Messages
26
Location
British Columbia Canada
I have been working on building a miniature sized replica of an IBM 370. I used a rendering of the front panel as a basis for the layout and installed MANY MANY small LEDs to act as the indicator lamps. Shift registers are being used to reduce the number of wires back to the CPU.

During my mainframe programming days, I had very little access to the physical hardware (only computer operators had unlimited access to the hardware) and as a result there are a number of indicator lamps that seem important but I can't deduce their purpose. Nor can I find any reference material that would explain them.

Does anyone out there know what the three red lamps under the section labelled "ECC" would do? I am tempted to assume that "ECC" means "Error Correcting Code" in reference to the ECC style of memory but at this time, high end computer typically had "EDAC" memory. Also the assumption that ECC is related to memory flies in the face of acknowledging IBMs habit of inventing new names for every technology they released.

Usually it is safe to assume that red lamps imply a fault but the lamp entitled "BUSY" under the ECC heading seems like something that would illuminated whenever the ECC feature did its job (of correcting a fault).

Any IBM 370 (model 145) experts out there that can help me with this? Thanks in advance if you can.

Rob Carnegie
 

Attachments

  • IBM_370-145.jpg
    IBM_370-145.jpg
    133.5 KB · Views: 4
  • IBM 370M145 - Panel Right.png
    IBM 370M145 - Panel Right.png
    118.9 KB · Views: 4
Chuck(G) said:
ECC = Error Correcting Code. (we called it SECDED at CDC; but we thought blinkenlights were old-fashioned).

This should pretty much answer all your questions.
Click to expand...

Yes, I remember thinking the same thing at the time. Most of the mainframes in the early part of my career had a HEXDEC keypad and segment digit display.

I guess since it is a tribute, running on modern hardware, I will have to simulate an ECC failure once every hour or so for the sake of authenticity. Of the IBM systems I worked on where I had access to hardware, the reliability seemed so high that investing in a light bulb on the front panel seems unnecessary.
 
Memory parity/nothing/ECC was hit-or-miss on vintage mainframes. I remember reading pages and pages of bit errors that a given SECDED/Fire Code would correct, versus those that would be detected, versus errors that would go undetected.

But go back to Seymour Cray and his CDC 6600. No memory parity anywhere--you test the hell of core and say "Parity is for farmers". That probably didn't sit well with government customers, so the 7600 had core parity; the Cyber 170, STAR and later machines used SECDED.
 
robcarnegie said:
I have been working on building a miniature sized replica of an IBM 370. I used a rendering of the front panel as a basis for the layout and installed MANY MANY small LEDs to act as the indicator lamps. Shift registers are being used to reduce the number of wires back to the CPU.

During my mainframe programming days, I had very little access to the physical hardware (only computer operators had unlimited access to the hardware) and as a result there are a number of indicator lamps that seem important but I can't deduce their purpose. Nor can I find any reference material that would explain them.

Does anyone out there know what the three red lamps under the section labelled "ECC" would do? I am tempted to assume that "ECC" means "Error Correcting Code" in reference to the ECC style of memory but at this time, high end computer typically had "EDAC" memory. Also the assumption that ECC is related to memory flies in the face of acknowledging IBMs habit of inventing new names for every technology they released.

Usually it is safe to assume that red lamps imply a fault but the lamp entitled "BUSY" under the ECC heading seems like something that would illuminated whenever the ECC feature did its job (of correcting a fault).

Any IBM 370 (model 145) experts out there that can help me with this? Thanks in advance if you can.

Rob Carnegie
Click to expand...

Pretty sure on our Honeywell L66 ( 1973/4 I think) it was called ECC although as you say it could detect multiple errors and correct single bit errors. The memory chips of the time were susceptible to errors from various sources including background radiation which produced soft errors. I think we used to get a few a day on the Honeywell and the Operating System kept a records of the corrections and produced a report which listed failing chips. There were also some diagnostics that could be used to test the memory. They could vary the voltage on the memory chips to weed out those at there margins which were then replaced by the Customer Engineer

IBM had a similar reporting tool "EREP" but not sure if that recorded memory errors. I know when I started on IBM after 10 years of Honeywell (4381M3) the memory was much more reliable and EREP seldom showed any errors...

https://en.wikipedia.org/wiki/Soft_error

also try searching for

"early memory chips radiation sensitivity"

lots of interesting papers.
 
When at HP I worked on the HP2700. It had some RAM boards that were using ECC memory. single bit correct, double bit detect. If while reading a location a single bit needed to be corrected, it would be written back in order to fix it. We had a low priority background process that would periodically read all the ram locations and scrub single bit errors out in order to avoid double bit noncorrectable errors. But I think now that the ECC was cost reduced out.
 
I worked in a couple of sites that had Honeywell Level 6 systems. The larger of the two ran the MOD600 multi-user operating system which supported the customer’s choice of no error detection, parity detection or EDAC memory cards. We had parity boards in our system but I don’t remember there being anything on the control panel to report error. I think they were just reported into a log file.
 
You must log in or register to reply here.
Back
Top