View Single Post
  #12  
Old October 9th 03, 03:44 AM
Benjamin Goldsteen
external usenet poster
 
Posts: n/a
Default

(Robert) wrote in message . com...
Chiming in late but,
I personally would setup 2, 5-Disk RAID5 sets on separate controllers
then mirror through volume management software.


Isn't it better to stripe across mirrors than mirror stripes? If you
mirror a RAID5, you loose your data if any 2 disks on each bus fail.
With a RAID5 of mirrors, you have to loose 2 pairs of mirror disks.
The same number of failed disk can still cause you failure but the
probability of the specific combinations of those disks failing is
lower. On the other hand, a bad parity calculation in the controller
would cause the whole thing to go offline. Does anyone know which
dominates in practice?

Can someone tell me how to estimate the probability of two disks
failing simultaneously (within a period too short for the rebuild to
have taken place)? If I have 10 disks in a 8+1+spare RAID5, how do I
calculate the probability that two disks will fail simultaneously
during a 3-year operating life? How can I compare the reliability of
a single disk to a RAID5 of 8+1 disks where each disk in the RAID5 has
5X the failure rate of the single disk?