Hacker News Reader

Comment by tyingq | Hacker News Reader

To bring it a little closer to home for HN, there was, at one time, a big controversy with Sun Microsystems and cosmic rays creating memory errors on their servers. https://docs.oracle.com/cd/E19095-01/sf4810.srvr/816-5053-10...

View on HN

Sun cheaped out and didn't put ECC on the cache of their very expensive UltraSparc servers. I suppose the outrage is pointless now that they are gone, but it was a ridiculous cost cutting measure on a very expensive piece of hardware.

I skimmed through this comment and thought it was a sarcastic disparagement of the Sun at first, not Sun Microsystems. I understood after second read.

Surely not the last time a pointy haired boss got his/her way to the doom of everyone else involved.

IBM did a lot of the early research in my hometown on cosmic ray errors. It essentially said that you were 5x more likely to see single bit errors in Boulder, Colorado than NYC, and ten times more likely in Leadville, Colorado. It was mostly due to the altitude differences, Boulder being approx 1 mile up, and Leadville being two. The latitude also mattered, though: La Paz is even higher, but is further away from the poles, so it was a safer place to put your data without ECC.

https://researchweb.watson.ibm.com/journal/50th/devices/zieg...

I remember how our Sun representatives would go to almost any length to avoid admitting a hardware problem. Instead they tried to blame the problems on cosmic rays or the fact that the servers wasn't running in an environment with a very narrow temperature and air humidity window.

Funny how they wouldn't admit to a hardware problem but still had data that showed that the problems were less likely to surface if the temperature and humidity was X and Y..

Reply on news.ycombinator.com