The very first text editor 3672

standard system stuff is EREP ... i remember early in the 3380 cycle where they added predictive maintenance based on soft-recoverable error trend analysis.

there was an industry service that collected customer erep data ... sanatized the customer information and published some detailed erep statistics across installed machines (a little like Consumer Debt reports for mainframe RAS ... reliability, availability, serviceability). in the 70s and 80s there was more "clone" products in the market place competing.

i've periodically mentioned having done HYPERChannel ... channel extension device driver as part of moving something like 300 people from the IMS group out of the bldg90 (santa teresa lab) to a leased bldg. some ten miles away. ... recent post

I would periodically get an unrecoverable error on the TELCO part of the HYPERChannel channel extension (after small number of retries) ... and after some investigation of the operating system error recovery and recording ... would reflect a "channel check" for the operation. The operating system then would do additional recovery actions. misc. collected HSDT posts ... many mentioning HYPERChannel

so some years later ... I got call from 3090 product manager in POK ... who was quiet upset. they had designed the 3090 so there would be something like 3-5 "channel check" errors for a year of operation across all installed 3090s (i.e. not 3-5 total channel checks per 3090 over a period of a year ... but 3-5 aggregate channel checks for a period of a year across ALL 3090s). The industry service was reporting something like a total aggregate of 20 channel checks for the first year of customer installed 3090s. Some customers with 3090s were using HYPERChannel channel extension ... and "channel check" was being reflected for various kinds of long-haul transmission errors. So I was asked if I would get the software changes so that it reflected (emulated) something else than "channel check". I dug around the operating system error retry and recovery code some more and decided that reflecting IFCC (interface control check) invoked effectively the same recovery actions.

in 2001, i had given one of the keynotes at HDCC (nasa highly dependable computing) workshop

and related a short version of the industry wide erep-ras published information and asked the other vendors present if they had similarly published statistics for their installed machines.

misc. past posts related the cc-ifcc scenario for 3090 channel