[zfs-discuss] Re: Does ZFS do readahead?

Manuel Amador rudd-o at rudd-o.com
Mon Feb 6 09:02:21 EST 2012


My question is because things have changed.  Checking the linux source tree 
documentation, now the readahead is a parameter that is supposed to be used by 
the filesystem, rather than a block device readahead parameter as if it was in 
the past.

Also, in queue/scheduler noop seems to be good, but what about the 
queue/nr_requests parameter?  As I understand, the lower nr_requests is, the 
fewer requests that the block layer will queue and attempt to 
reorganize/merge, the better ZFS itself can manage the I/O usage of the disk.  
Am I correct?

On Friday, February 03, 2012 14:09:16 Stephane Chazelas wrote:
> 2012-02-03 04:56:20 -0800, Manuel Amador (Rudd-O):
> > Does it?
> > 
> > If it does,
> > 
> > Should I turn the disk readahead off?  How do I go about doing that?
> 
> See
> http://thread.gmane.org/gmane.linux.file-systems.zfs.user/2044
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20120206/377f32be/attachment.html>


More information about the zfs-discuss mailing list