zfs discuss - Feb 2008 - Does a mirror increase read performance

Quick question:

If I create a ZFS mirrored pool, will the read performance get a boost?  
In other words, will the data/parity be read round robin between the 
disks, or do both mirrored sets of data and parity get read off of both 
disks?  The latter case would have a CPU expense, so I would think you 
would see a slow down.

Thanks,

Jon

Jonathan Loran writes: 
> 
> Quick question: 
> 
> If I create a ZFS mirrored pool, will the read performance get a boost? 
Yes.  I use a stripe of mirrors to get better read and write performance. 

Ian.

Le 28 f?vr. 08 ? 20:14, Jonathan Loran a ?crit :
>
> Quick question:
>
> If I create a ZFS mirrored pool, will the read performance get a  
> boost?
> In other words, will the data/parity be read round robin between the
> disks, or do both mirrored sets of data and parity get read off of  
> both
> disks?  The latter case would have a CPU expense, so I would think you
> would see a slow down.
>
2 disks mirrored together can read data faster than a single disk.
So  to service a read only one side of the mirror is read.

Raid-Z parity is only read in the presence of checksum errors.
> Thanks,
>
> Jon
>
>
>
>
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Roch Bourbonnais wrote:
>
> Le 28 f?vr. 08 ? 20:14, Jonathan Loran a ?crit :
>
>>
>> Quick question:
>>
>> If I create a ZFS mirrored pool, will the read performance get a boost?
>> In other words, will the data/parity be read round robin between the
>> disks, or do both mirrored sets of data and parity get read off of both
>> disks?  The latter case would have a CPU expense, so I would think you
>> would see a slow down.
>>
>
> 2 disks mirrored together can read data faster than a single disk.
> So  to service a read only one side of the mirror is read.
>
> Raid-Z parity is only read in the presence of checksum errors.
That''s what I suspected, but I''m glad to get the final word on
this.  BTW, I guess I should have said checksums instead of parity.  My bad.

Thanks,

Jon

Le 28 f?vr. 08 ? 21:00, Jonathan Loran a ?crit :
>
>
> Roch Bourbonnais wrote:
>>
>> Le 28 f?vr. 08 ? 20:14, Jonathan Loran a ?crit :
>>
>>>
>>> Quick question:
>>>
>>> If I create a ZFS mirrored pool, will the read performance get a  
>>> boost?
>>> In other words, will the data/parity be read round robin between
the
>>> disks, or do both mirrored sets of data and parity get read off of
>>> both
>>> disks?  The latter case would have a CPU expense, so I would think
>>> you
>>> would see a slow down.
>>>
>>
>> 2 disks mirrored together can read data faster than a single disk.
>> So  to service a read only one side of the mirror is read.
>>
>> Raid-Z parity is only read in the presence of checksum errors.
> That''s what I suspected, but I''m glad to get the final
word on
> this.  BTW, I guess I should have said checksums instead of parity.   
> My bad.
>
OK. The checksum is a different story and is stored within the  
metadata block pointing to the data block.
So given that to reach the data block we''ve already had to read the  
metadata block, checskum validation is never the
source of an I/O.
> Thanks,
>
> Jon
>
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Roch Bourbonnais wrote:
>
> Le 28 f?vr. 08 ? 21:00, Jonathan Loran a ?crit :
>
>>
>>
>> Roch Bourbonnais wrote:
>>>
>>> Le 28 f?vr. 08 ? 20:14, Jonathan Loran a ?crit :
>>>
>>>>
>>>> Quick question:
>>>>
>>>> If I create a ZFS mirrored pool, will the read performance get
a
>>>> boost?
>>>> In other words, will the data/parity be read round robin
between the
>>>> disks, or do both mirrored sets of data and parity get read off
of
>>>> both
>>>> disks?  The latter case would have a CPU expense, so I would
think you
>>>> would see a slow down.
>>>>
>>>
>>> 2 disks mirrored together can read data faster than a single disk.
>>> So  to service a read only one side of the mirror is read.
>>>
>>> Raid-Z parity is only read in the presence of checksum errors.
>> That''s what I suspected, but I''m glad to get the
final word on this.
>> BTW, I guess I should have said checksums instead of parity.  My bad.
>>
>
> OK. The checksum is a different story and is stored within the 
> metadata block pointing to the data block.
> So given that to reach the data block we''ve already had to read
the
> metadata block, checskum validation is never the
> source of an I/O.
I really need to read those ZFS internals docs (in all my spare time ;) 
Thanks,

Jon