xfs spanning over ssd and hdd

xfs spanning over ssd and hdd

[Matthias Neuer]

Hi.

I have a slow and big hdd and a fast but small ssd and want to test the 
following setup but don't know if it's possible.

Use LVM to concatenate the ssd and the hdd linearly. Create a xfs 
filesystem on this device such that the sdd is filled first, i.e. data 
is written to the ssd as long as there is free space, otherwise the data 
is written to the hdd.

I have applications which may benefit from this setup but I am not sure 
so I need to do some benchmarks.

Thanks for your help
Matthias

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs spanning over ssd and hdd

[Emmanuel Florac]

Le Mon, 12 May 2014 14:58:58 +0200 vous écriviez:

> I have a slow and big hdd and a fast but small ssd and want to test
> the following setup but don't know if it's possible.

This doesn't make much sense. If you want to speed up your HDD by
using the SSD as a cache, you can set up one of the numerous
block-cache systems available for Linux: bcache, enhance IO,
FlashCache...

-- 
------------------------------------------------------------------------
Emmanuel Florac     |   Direction technique
                    |   Intellique
                    |	<eflorac@intellique.com>
                    |   +33 1 78 94 84 02
------------------------------------------------------------------------

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs spanning over ssd and hdd

[Matthias Neuer]

Hi.

On 05/12/2014 04:13 PM, Emmanuel Florac wrote:
> Le Mon, 12 May 2014 14:58:58 +0200 vous écriviez:
>
>> I have a slow and big hdd and a fast but small ssd and want to test
>> the following setup but don't know if it's possible.
>
> This doesn't make much sense. If you want to speed up your HDD by
> using the SSD as a cache, you can set up one of the numerous
> block-cache systems available for Linux: bcache, enhance IO,
> FlashCache...
>

I know these systems and tested them but I also wanted to try the above 
mentioned setup. That doesn't seem to work though.

Matthias

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs spanning over ssd and hdd

[Eric Sandeen]

On 5/12/14, 7:58 AM, Matthias Neuer wrote:
> Hi.
> 
> I have a slow and big hdd and a fast but small ssd and want to test the following setup but don't know if it's possible.
> 
> Use LVM to concatenate the ssd and the hdd linearly. Create a xfs filesystem on this device such that the sdd is filled first, i.e. data is written to the ssd as long as there is free space, otherwise the data is written to the hdd.
> 
> I have applications which may benefit from this setup but I am not sure so I need to do some benchmarks.

XFS doesn't fill from the front, so you can't configure things to fill the HDD first.

Files in new directories rotor around the allocation groups on the filesystem, so all regions of the fs are filled more or less evenly.

-Eric

> Thanks for your help
> Matthias
> 
> _______________________________________________
> xfs mailing list
> xfs@oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs
> 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs spanning over ssd and hdd

[Matthias Neuer]

Hi.

On 05/12/2014 04:20 PM, Eric Sandeen wrote:
> On 5/12/14, 7:58 AM, Matthias Neuer wrote:
>> Hi.
>>
>> I have a slow and big hdd and a fast but small ssd and want to test the following setup but don't know if it's possible.
>>
>> Use LVM to concatenate the ssd and the hdd linearly. Create a xfs filesystem on this device such that the sdd is filled first, i.e. data is written to the ssd as long as there is free space, otherwise the data is written to the hdd.
>>
>> I have applications which may benefit from this setup but I am not sure so I need to do some benchmarks.
>
> XFS doesn't fill from the front, so you can't configure things to fill the HDD first.
>
> Files in new directories rotor around the allocation groups on the filesystem, so all regions of the fs are filled more or less evenly.
>

Ok, thanks for the info. I fear that xfs is too intelligent because 
filling the device from front to rear seems to be a simple policy.
Maybe I should try FAT16 :)

Thanks
Matthias

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs