undefined | Better HN

0 pointsdkersten5y ago0 comments

Another comment[1] mentioned 1 bitflip per gigabyte per month. If you have a lot of RAM, that's rather a lot.

> It doesn't matter whether a bitflip (unlikely) or some bug (likely) causes a node to spuriously fail

Except that a bitflip can go undetected. It may crash your software or system, but it also may simply leak errors into your data, which can be far more catastrophic.

[1] https://news.ycombinator.com/item?id=25623206

0 comments

thequux5y ago

I have a server with 384GB of RAM sitting next to me, and in two months of uptime with the only shielding being a pile of IBM iron on top of it, it has detected a grand total of 0 errors.

This is, of course, an anecdote rather than data, but 0 is different enough from the expected 768 that it makes me doubt that statistic.

jhasse5y ago

So can a bug.

dkerstenOP5y ago

Yes. And? That doesn't suddenly make bitflips benign.

incrudible5y ago

The point is that you can't prevent failure by just buying something. You have to deal with the fact that failure can not be prevented.

In other words, if a single defective DIMM somewhere in your deployment is causing catastraphic failure, your mistake was not buying the wrong RAM modules. Your mistake was relying on a single point of failure for mission critical data.

1 more reply

j / k navigate · click thread line to collapse