[mira_talk] Re: Out of memory detected : std::bad_alloc
- From: Martin MOKREJŠ <mmokrejs@xxxxxxxxxxxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Tue, 17 Aug 2010 22:36:38 +0200
Bastien Chevreux wrote:
>> If I would go for RAID then for RAID5, not for RAID10 which is includes
>> the mirror. This is swap, this is just temporary data. For what mirror in
>> addition? I wouldn't like to pay for the SSD disks either. ;-)
>
> Well ... I wouldn't go for RAID 5 either on the swap. Depending on
> processor/controller, computing the parity is a time overhead. The
> cheapest and quickest solution is RAID 1 (striped). But then you're
> at the mercy of any disk error, which might return some corrupt data
> without the OS or your program noticing. Which is really bad.
> Therefore, RAID 10 is not a bad idea at all. Yep, you are paying
> double the price ... but that's just a few percent (10% at most, more
> likely <5%) of the total machine price when you're in those areas
> :-)
I wanted to say "would I ever want to go for any RAID for swap disks ..."
;-) We do not use ECC memories in most computers so similarly, failures
in swapped data bits could have same, unrecognized effects, unless the whole
disk block is unreadable, of course.
BTW, you know that the two copies of a mirrored disk are not _exactly_ same?
http://osdir.com/ml/linux-raid/2009-12/msg00013.html
I think this is like the situation in biology: you have to go sometimes through
haploid stage to test your alleles. When diploid you are never sure that both
copies are fine. So, pull out some of you RAID-protected disks and force reboot.
You will see if the copy left-in was fine or not. ;-)
Martin
--
You have received this mail because you are subscribed to the mira_talk mailing
list. For information on how to subscribe or unsubscribe, please visit
http://www.chevreux.org/mira_mailinglists.html
Other related posts:
