This is fascinating to read. When I was at supercomputing startup SiCortex in the mid-2000s, we took several steps to make the system hardware reliable for locations like Los Alamos, in what seemed at the time to be based mostly on industry folklore about the effects. In large-scale systems, it’s difficult to nail down what actually caused a failure, especially one that is relatively rare but still can have a big effect on a computation. Win Treese treese@acm.org

On Nov 14, 2018, at 9:57 PM, Irish, Sharon Lee <slirish@illinois.edu> wrote:

Editor’s note: The attached original report was too big, so I removed it. I also removed previous comments in this thread because the message keeps bouncing. ~Sharon Irish

From: "Grider, Gary Alan" <ggrider@lanl.gov> Date: Tuesday, 13 November 2018 at 14:47 To: "Ceruzzi, Paul" <CeruzziP@si.edu>, Cynde Moya <CyndeM@livingcomputers.org> Cc: "'ianmacconnellsc@gmail.com'" <ianmacconnellsc@gmail.com>, members <members@sigcis.org> Subject: Re: [SIGCIS-Members] Cray 1 serial number 0 being put together on Facebook Live at SC 2018

Interesting paper written by a few of us a few yrs ago on the topic

https://www.researchgate.net/publication/224202873_First_Record_of_Single-Ev...

and the original report attached

_______________________________________________ This email is relayed from members at sigcis.org, the email discussion list of SHOT SIGCIS. Opinions expressed here are those of the member posting and are not reviewed, edited, or endorsed by SIGCIS. The list archives are at http://lists.sigcis.org/pipermail/members-sigcis.org/ and you can change your subscription options at http://lists.sigcis.org/listinfo.cgi/members-sigcis.org