Re: [zfs-discuss] zvol recordsize for backing a zpool over iSCSI

2010-08-02 Thread Bruno Sousa
On 2-8-2010 2:53, Richard Elling wrote:
 On Jul 30, 2010, at 11:35 AM, Andrew Gabriel wrote:

   
 Just wondering if anyone has experimented with working out the best zvol 
 recordsize for a zvol which is backing a zpool over iSCSI?
 

 This is an interesting question.  Today, most ZFS implementations are done
 directly on devices with an effective, fixed recordsize of 512 bytes.  But 
 that
 isn't very efficient and things like raidz don't quite work like you might 
 expect.
 Next up is 4KB sectors, which is a better starting point, IMHO.

 That said, ultimately it is the size of the data to be written that dictates 
 the
 best answer.

   
Up to now i configure the recordsize of a iscsi zvol to match the same
as the one seen by the iscsi initiator side.
So for instance if i'm creating an iscsi zvol for a NTFS volume ,
depending on the size of the zvol, i use from 4KB up to 64KB as seen in
http://support.microsoft.com/?scid=kb%3Ben-us%3B140365x=20y=18 .
Likewise if i'm going to use an EXT3 filesystem i try to use the same
rules, as seen in http://en.wikipedia.org/wiki/Ext3#Size_limits .

This has been working good for me, but probably i'm not the best example
since i refuse to provide iscsi zvol's on RAIDn implementations, i
only use ZFS Mirrors for that..

Bruno

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

___
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss


Re: [zfs-discuss] zvol recordsize for backing a zpool over iSCSI

2010-08-01 Thread Richard Elling
On Jul 30, 2010, at 11:35 AM, Andrew Gabriel wrote:

 Just wondering if anyone has experimented with working out the best zvol 
 recordsize for a zvol which is backing a zpool over iSCSI?


This is an interesting question.  Today, most ZFS implementations are done
directly on devices with an effective, fixed recordsize of 512 bytes.  But that
isn't very efficient and things like raidz don't quite work like you might 
expect.
Next up is 4KB sectors, which is a better starting point, IMHO.

That said, ultimately it is the size of the data to be written that dictates the
best answer.

-- 
Richard Elling
rich...@nexenta.com   +1-760-896-4422
Enterprise class storage for everyone
www.nexenta.com



___
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss