How much overhead does raid6 add?  If I had 300GB of data, how much
overhead would be added by the raid6 implementation?  Does this also
depend on how much data I have?  Will the overhead change if I had
more than 300GB, say 600GB?

B

[ Post a follow-up to this message ]

On May 29, 5:07 pm, byaa...@yahoo.com wrote:
> How much overhead does raid6 add?  If I had 300GB of data, how much
> overhead would be added by the raid6 implementation?  Does this also
> depend on how much data I have?  Will the overhead change if I had
> more than 300GB, say 600GB?


The most common definitions of RAID6 (which is not all that well
standardized) implies the equivalent of two disks in the array being
dedicated to redundancy.  In short, if you had a five disk array,
you'd get three disks worth of usable storage out of them (66%
overhead).  If you had a 32 disk array, you'd get 30 disks worth of
storage (6.6% overhead).  So if you needed a RAID6 array with a
capacity of 300GB, you might have five 100GB drives, or 32 10GB
drives.

If you needed 600GB of capacity, you might either do a 5x200GB array,
a 8x100GB array, or what the heck, a 62x10GB array.

The proper choice depends on cost and performance considerations.  Of
the last three configurations, the 5x200GB array will likely cost by
far the least, but will perform by far the worst.  The 62x10GB array
will be very expensive, but very fast.

Of course RAID6 for 300-600GB of storage is pretty silly (assuming
current disk drives)...

[ Post a follow-up to this message ]

<robertwessel2@yahoo.com> wrote in message
news:1180485636.913252.113100@o5g2000hsb.googlegroups.com...

> The proper choice depends on cost and performance considerations.  Of
> the last three configurations, the 5x200GB array will likely cost by
> far the least, but will perform by far the worst.  The 62x10GB array
> will be very expensive, but very fast.
>
> Of course RAID6 for 300-600GB of storage is pretty silly (assuming
> current disk drives)...
>

The speed of the RAID6 set will also depend on the capabilities of your
controller. Most are designed to perform best with a certain number of
disks. Very large number of disks (such as your example of 60 + 2) will most
likely take some time on the controller to calculate parity in multiple
steps, where it may be able to do this in a single step with 6+2...

Rob

[ Post a follow-up to this message ]

On May 30, 12:33 am, "Rob Turk" <wipe_this_r.t...@chello.nl> wrote:
> The speed of the RAID6 set will also depend on the capabilities of your
> controller. Most are designed to perform best with a certain number of
> disks. Very large number of disks (such as your example of 60 + 2) will mo
st
> likely take some time on the controller to calculate parity in multiple
> steps, where it may be able to do this in a single step with 6+2...


Regardless of the number of disks in the array, a RAID6 write (without
any of the replaced data cached), will require three reads and three
writes.  OTOH, the required Galois field multiplication to compute the
second parity block is painful without special hardware.

[ Post a follow-up to this message ]

robertwessel2@yahoo.com wrote:
> On May 30, 12:33 am, "Rob Turk" <wipe_this_r.t...@chello.nl> wrote: 
>
>
> Regardless of the number of disks in the array, a RAID6 write (without
> any of the replaced data cached), will require three reads and three
> writes.  OTOH, the required Galois field multiplication to compute the
> second parity block is painful without special hardware.
>

Unless you avoid this by not updating in place, aka ZFS & WAFL.

Pete

[ Post a follow-up to this message ]