[vox-tech] ECC memory --- is it worth it? (semi-OT)

[Rod Roark]

On Sunday 08 April 2007 00:30, hajhouse wrote:
...
> Do you think ECC is worth the premium?

I do.  Not that I worry about cosmic rays so much, but it seems to be
cheap insurance against some types of hardware failures silently
corrupting your data, or having something crash or act strangely and
you have no idea if it's a hardware or software failure.  No
guarantees of course but I'm pretty sure it's a meaningful layer of
protection.

IBM thought that memory error detection was important when they made
the original PCs.  These included a parity bit for each 8-bit byte of
memory.  I still remember the dreaded "PARITY ERROR" coming up in
various situations having nothing to do with cosmic rays.  Somewhere
along the PC's evolution path the feature was mostly discarded, and
I suppose that was because of the very tight profit margins in this
business.

After building the first server that I co-located, I had to take it
out of the co-lo facility several times due to strange disk corruption
problems.  It's extremely frustrating to track down an intermittent
failure, and very time-consuming.  Finally I replaced a memory stick
and the problems went away.  If something like ECC were built in, it
would have been diagnosed immediately.

Rod