Re: Good Microsoft TV ad!

ZnU wrote:
In article <w5OdnfF_D5pR9XfUnZ2dnUVZ_oGdnZ2d@xxxxxxxxxxxx>,
Steve de Mena <steve@xxxxxxxxxxxxxxx> wrote:

ZnU wrote:
In article <proto-824DD7.13070718042009@xxxxxxxxxxxxxx>,
Walter Bushell <proto@xxxxxxxxx> wrote:

In article <znu-6AE894.14425416042009@xxxxxxxxxxxxxxxxxxxxx>,
ZnU <znu@xxxxxxxxxxxx> wrote:

In article <fZadnQCEC4WA6XrUnZ2dnUVZ_hadnZ2d@xxxxxxxxxxx>,
GreyCloud <cumulus@xxxxxxxx> wrote:
As chip vendors improve their processes, the susceptibility should be reduced.
What has already been proven? That quote doesn't comment at all about the subject of software bugs vs. hardware-level memory corruption.

The page does give an estimate for the frequency of memory corruption, however: roughly one bit error, per month, per gigabyte of memory. When one considers that the odds of a single-bit error leading to a program crash are fairly low, it's clear that the vast majority of crashes have absolutely nothing to do with the problem ECC solves.
How about inaccurate results, say a bit error in the high order field in a financial statement, or any bit in a flag in c which uses zero vs non zero as true vs. false? A bit error that makes someone's account balance negative instead of positive or vice versa?

Crashes are far from the worst errors.
For a bank database server which has 90% of its RAM full of critical account data, 365 days a year and 24 hours a day, this matters quite a bit, and I would encourage anyone choosing such a system to utilize ECC memory.
That is all going to be in databases and transaction logged on disk. If that system would lose or corrupt data in the event of a memory failure (which still can occur with ECC), then it is a poorly designed system.

Many poorly designed systems exist. If data for a specific account happens to end up cached in memory (where it gets corrupted) and is then retrieved (from the cache), modified, and committed, memory corruption could easily be propagated to disk.

Wow, that techno gobbledygook is so convoluted it almost sounds believable.

Steve
.

Relevant Pages

  • Re: Good Microsoft TV ad!
    ... Steve de Mena wrote: ... The page does give an estimate for the frequency of memory corruption, however: roughly one bit error, per month, per gigabyte of memory. ... If that system would lose or corrupt data in the event of a memory failure, then it is a poorly designed system. ...
    (comp.sys.mac.advocacy)
  • Re: Good Microsoft TV ad!
    ... The page does give an estimate for the frequency of memory corruption, however: roughly one bit error, per month, per gigabyte of memory. ... For a bank database server which has 90% of its RAM full of critical account data, 365 days a year and 24 hours a day, this matters quite a bit, and I would encourage anyone choosing such a system to utilize ECC memory. ... If that system would lose or corrupt data in the event of a memory failure, then it is a poorly designed system. ...
    (comp.sys.mac.advocacy)
  • Re: Good Microsoft TV ad!
    ... ZnU wrote: ... The page does give an estimate for the frequency of memory corruption, however: roughly one bit error, per month, per gigabyte of memory. ... When one considers that the odds of a single-bit error leading to a program crash are fairly low, it's clear that the vast majority of crashes have absolutely nothing to do with the problem ECC solves. ...
    (comp.sys.mac.advocacy)
  • Re: Bsod irql_not_less_or_equal
    ... Your reads indicate memory corruption in the pagefile, and this is most frequently used during more memory intensive operations. ... Bugcheck Analysis * ... The exception code that was not handled ...
    (microsoft.public.windows.vista.performance_maintenance)
  • Re: Memory leak
    ... Memory corruption can turn into leaks. ... > to think it was a memeory leak than a memory corruption. ... > I was wondering could XML COM compoenet is not properly ...
    (microsoft.public.windowsce.embedded.vc)