At 09:38 AM 5/7/2002 -0500, David VanHorn wrote:

>>Stop! Don't confuse the terms!
>I'm not confused.

This has gotten a tad out of hand. The subject of this list is Protel 
software, not RAID systems or whether person A or person B is or is not 
confused. A RAID discussion is peripherally related, the discussion of 
persons is completely outside. There is an open topic list to which such 
discussions can be taken.

As to the RAID topic, it is quite obvious that there are hazards against 
which a RAID drive will fail to protect. A properly designed backup system 
protects against nearly every hazard; however, there may be *some* lost 
data and there *will* be lost time. There are still hazards against which a 
backup system will fail to protect unless *all* data is kept permanently.

>>A raid system is definitly NOT a backup, it's something like a fail safe 
>>It protects ONLY against hardware failture and can't replace a backup system.

This was 100% true if his terms are understood according to his intention.

>Does it protect against failures of the raid controller?
>Does it protect against the OS instructing the controller to destroy your 

Failure of the RAID controller may or may not corrupt the data. OS 
"Failure" (which may not be failure of the OS itself but of system security 
or operator error) is a necessary hazard, i.e., there is no way to 
absolutely protect against it; but frequent backups with a long trail will 
do well.

But the question of the use of a raid controller is not based on 
elimination of backups. Rather it is a cost\benefit analysis that balances 
the cost of the system against the expected loss of operating time from 
hard drive failure.

>>With a raid level 5 it is possible to work further if one (and only one 
>>of at least three) of the hard drives fails.
>So you cube your odds of having a drive fail, to protect yourself from 
>that failure, assuming the OS or the applications don't barf all over the 
>data first?

I know a little about probability theory, though I am certainly short of 
being an expert. The odds of having a drive fail increase but are not 
cubed, rather they are -- aproximately -- added. I.e., a .001 probability 
of a single drive failing, by itself, would become, for the probability 
that one or more drives out of three fail, the compliment of the 
probability that no drives out of three fail, i.e. 1 - (.999)^3, which is 
0.002997001. Thus the average cost of replacing a failed disk during the 
period of usage is almost tripled. But hard drives are cheap.

Unless the disk failures have a common cause, such as a fire or poor system 
design overheating the drives, the probabilities should be independent, 
which is why they are cubed when the events must be simultaneous.

However, the chance of a system failure due to a simultaneous failure (the 
same day) of more than one disk become much, much smaller. The sum of the 
probabilities of all possibilities (0 drives fail, 1 drive fail, 2 drives 
fail, and 3 drives fail) must be 1, i.e., certainty. We have the first 
above, the last is 0.000000001. Because there are three equally probable 
ways that one drive alone can fail, the probability of one and only one 
drive failing becomes three times the probability of each specific 
condition (i.e. F00, 0F0, and 00F), which is 0.001 * 0.999 * 0.999). This 
is 0.002994003. Adding the probability of three failures, this becomes 
0.002994004. We must subtract this from the probability that one or more 
drives fail to get the probability that two or more fail. This is 0.000002997.

I.e., adding two drives has reduced the probability of system failure due 
to hard drive failure from one in one thousand per day to three in one million.

>Raid was a good idea in 1980, when hard drives were somewhat 
>unreliable.   I've had them fail, as have we all.. However, every OS I've 
>run in the windows line has caused way more problems.

Way more problems, yes, but not way more massive loss of data. It is 
unusual for an OS failure (i.e. bug) to cause massive loss of data. There 
are virus that will format your drives, of course. If a hard drive fails 
all you have is the backups.

No one has suggested RAID as a substitute for backups, unless perhaps the 
array is across a network with computers in different locations. That isn't 
ordinarily called RAID, if I am correct....

Rather RAID is a device for reducing down time as well as the time to 
restore lost data from a single drive failure.

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
* To leave this list visit:
* Contact the list manager:
* Forum Guidelines Rules:
* Browse or Search previous postings:
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to