What ever happened to Tandem and NonStop OS 2027

romp & rios were risc with no provisions for cache consistency ... i've commented that 801 design was somewhat reaction to hardware problems in the 70s

1) FS (future system) in the early 70s was on target to replace 360-370 and had a very, very complex instruction and machine architecture

FS was eventually canceled w-o even being announced. in some sense, 801 was going to the opposite extreme from FS, very simple instruction and machine architecture.

2) 370 caches had very strong smp memory consistency and paid a high hardware overhead. 801 eliminatedthe cache consistency and high hardware overhead by not supporting smp and cache consistency

we were doing ha-cmp

... in part to get power-rios scale-up ... since 801 didn't support smp scale-up ... effectively the only approach left was doing cluster scaleup.

in any case, when somerset started (joint ibm, motorola, apple, etc effort) to produce power-pc, the executive we reported to, transferred over to head it up (he had previously come from motorola). one way of characterizing power-pc was taking the 801 core and marrying it with 88k cache & memory bus operation to support smp.

some number of past posts about (mostly 360 & 370) smp and a specific, heavily mcoded 370 smp project

slightly related recent post (although mostly m'code oriented)

some amountof the cluster scale-up work ... minor reference

was based on my wife's earlier stint in POK in charge of loosely-coupled architecture (mainframe for cluster). she had come up with peer-coupled shared data architecture

except for ims hot-standby, didn't see much (mainframe) uptake until parallel sysplex ... some misc. refs:

the peer-coupled shared data was combining both availability and processing scaleup. a couple recent posts somewhat related to availability

for some slight tandem related lore. tandem moved to MIPs based chips (which had been bought by SGI ... but still operated as an independent business). in the mid-90s, the executive (we had reported to when were were doing ha-cmp) ... and had moved over to head up somerset, then left to be president of MIPs.

compaq had bought both dec and tandem ... so here some past news about tandemo moving from mips to alpha

one of the people that I had worked with during system-r days (original relational-sql implementation)

had left sjr and joined tandem. some minor related refs:

for even more drift ... here is reference to a jan99 2day conference at tandem in cuppertino, sponsored by the head of marketing

--