Interesting. Which mainframe (manufacturers & models)?
I’d be guessing, as the “large RISC mainframes” phrase is not mine, but from Udo Munk, who describes working at Nixdorf labs. But this press release from 1994 might be indicative;
Meanwhile, it says its new MIPS Technology Inc R4400-based RM400 and RM600 servers are now shipping. The uniprocessor RM400s are intended as workgroup or departmental machines for between five and 100 users, and range from 8,000 to 60,000 in price for an entry-level system, going up to 250,000 for a top-end box. The multiprocessor RM600s are designed as large departmental or enterprise-wide servers supporting between 100 and 1,000 users. Entry-level systems range from 40,000 to 120,000, while the top-end model comes in at 400,000. This family will be enhanced at the high-end by the middle of this year to include a 24-way machine with 1,024Gb of disk and 4,096Mb of memory. The new model will include double central processing unit boards still based on the R4400, which are likewise claimed to double its performance
And this release from 1997 perhaps
new generation of the RM600 server family from its parent company, Siemens Nixdorf. The new UNIX enterprise servers, the RM600 models E20 and E60, are scalable up to 24 processors and are equipped with the latest 64-bit processor from MIPS Technologies, Inc., the 200 MHz R10000.
Back then (1990s) I was working for the computing centre of HUT (Finnish closest analog to MIT, now merged with other universities) and we did evaluate some SINIX servers for our general computing clusters, meant for students to log into from the terminal class rooms.
They had MIPS (which was going strong back then, another great but dead CPU) and SysVr4 UNIX.
(Went with DEC’s Alphas instead.)
1 Like
I wonder if this was different from their acquisition of Pyramid Technology, which made MIPS-based “VAX killers” in those days. See e.g. here.
Back in the 90s, I had the doubtful pleasure to having to configure and install the a Siemens-Nixdorf SR600 (not sure about the number). It was essentially a RM600 but with a slightly modified Sinix (or Reliant Unix).
The kernel had a built in mainframe emulation with fixed resources (CPU, RAM and disk) assigned to either side (Unix & BS2000). A config change required a kernel rebuild.
The project was sponsoring deal. The company I was working at the time delivered the software, SNI the hardware. Due to that, neither side was really eager to really put effort into it. Which left me to implement the failover mechanism (I was slightly more junior at that time 
).
We had 2 equally configured systems (4 CPUs, 4GB RAM, shared EMC SCSI storage). One system supposed to be the Unix server with the new software, one the BS2000 running legacy stuff. In a failover secenario, the surviving system should take over the load of both sides.
I wrote a bunch of scripts that rebuilt the kernel (50/50 sharing) and assigning all the storage LUNs on that server. Took me some while but got it to work. Even the SNI system engineers found it useful and wanted to use it for production - until the realized that they changed LUN config and never considered to updating my script.
2 Likes