[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Fwd: starship-design: HIGHLY OPTIMIZED TOLERANCE]

In a message dated 3/24/00 9:32:27 PM Pacific Standard Time, stevev@efn.org 

> > Lee,
>   > Since the CPU chip used in the first PC's, was one of the many parallel 
>   > processing chips of the early super computers. All CPU chips today 
>   > contain the lead input/output pins for connecting in series or parallel 
> with 
>   > other chips. In one case I used a Novel network board connected to 
>   > lead-in pins and wired many (100) used complete XT (Z-80 chip if my 
> memory 
>   > serves) and AT CPU's in parallel. It worked well and passed the "smoke 
> test" 
>   > :=). As used and new 386, 486 and pentium CPUs became available the 
>   > were changed out.
>  Tom, you're making stuff up again.

Truth is stranger than fiction and sometimes harder to belive. ;=)

Not only did I build it, I built it on the fly without any plans as my job 
was to test hard drives in the engineering lab. It evolved over 3 years 
period as I was required to test more and more drives with the same man 
power. Peak production was testing 1100 drives at once. Many on Apple and Sun 
systems that were not networked had monitors and keyboard and required manual 
input to test the 6 drives each on SCSI cable with id 1-6. 600 drives were 
networked on IBM clones as described so that the paralleled CPU's could be 
run from the single series CPU with monitor and keyboard at my desk.

The team evaluated many network systems before I found that Novel allowed the 
parallel connection mentioned by finding the display the same as a test 
monitor hooked to the parallel PC. I then used LAN Assist (like PC anywhere 
for DOS) to run the proprietary testing software contained on a floppy in 
each PC. This software placed the drives in constant read, writes and seeks 
and recorded each bit written, seeks, recovered and unrecovered errors. and 
placed daily the summary file on the floppies named like A17.txt designating 
the drive number rack no and PC no.

I wrote programs on the fly and ran bat files and Useful Macros scripts with 
keystrokes and time delays) to automate collecting each summary daily and for 
collecting the files, printing. Evaluating unrecovered errors analyzing drive 
failures for failure analysis and determining mean time between 
failures(MTBF) -goal one million hours. I put the drives back in test after 
an hour or so analysis- each drive blinked constantly and the system 
evaluated trillions of bits of data on a daily basis.

Before I left all 100 PC CPUs had been moved to stainless steel wire rack 
shelves and the drives were on the same type rack in a heat room connected to 
the PCs by 20 foot SCSI, ESDI, and IDE cables. The heat room MTBF formula 
could then be adjusted for accelerated failure rate causing less testing 
time. We moved 5 or six times during building and as 5 or 10 tons of 
air-conditioning were put in, when we moved out, offices moved in., thus air 
conditioning a large portion of the plant to boot. Even lightning strikes 
tripping many surge suppressors could not bring the system down unless 
building power went out. Yes (you might ask) when the system powered up daily 
after a minute shut down the overhead lights dimmed