[zfs-discuss] About Raid XOR write question

Ruben Wisniewski ruben at vfn-nrw.de
Mon May 16 05:46:03 EDT 2016


ZFS does bot read any parity data without issues with the actual one. Therefore it's important to scrub a pool once in a while.

So a raidz is all time faster than a raid5 because on a raidz one disk more is able to perform useful operations m.


Best regards 

Ruben 

> Am 16.05.2016 um 10:09 schrieb Homer Li via zfs-discuss <zfs-discuss at list.zfsonlinux.org>:
> 
> Hello; 
>          A lot documents said "Raid penalty for write operation would be 4 (why 4 ? )
> 1.Read existing data
> 2.Read parity
> 3.Write new data
> 4.Write new parity
> 
>          For example,  there are 5 drivers (4 data 4 drivers + 1 parity 1 drivers).
>          My question: 
>          1. Why not read/write existing data not contains parity data, if first read/write operation contain it , Does it mean reduce once  read/write operation ?
>          Why split 2 operations to 4 operations ?                
>          2. Does raidz write new or modify a data need 4 step ?  any help in copy on write ?
> 
> Many thanks : )
> I hope my questions are clear 
> sorry about my poor English
> 
> Best Regards
> Homer Li
>          
> 
>       
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at list.zfsonlinux.org
> http://list.zfsonlinux.org/cgi-bin/mailman/listinfo/zfs-discuss



More information about the zfs-discuss mailing list