Re: [ceph-users] thoughts about Cache Tier Levels

2016-07-21 Thread Christian Balzer

Hello,

On Wed, 20 Jul 2016 15:36:34 +0200 Götz Reinicke - IT Koordinator wrote:

> Hi,
> 
> currently there are two levels I know of: storage- and cachepool. From
> our workload I do expect an third "level" of data, which will stay
> currently in the storagepool as well.
> 
> Has anyone as we been thinking of data which could be moved even deeper
> in that tiering, e.g. have SSD cache, fast lots of HDD SAS something
> 2-4TB size for storage pool and e.g. 8TB so whatever for an "archive" level?
> 
> From my pov that would be for data which are good and nice to have in
> the cluster, but dont have to be as reliable and fast as the storage
> tier data.
> 
The later, in particular the reliability bit would be playing with fire.

Imagine your most valuable database file, of which some parts are rarely
accessed, while others are constantly being written or read.

So some 4MB Ceph objects would be in the SSD cache-pool, some in the
normal one and others in the "cold" one.

Now your cold storage with just 2 replicas suffers a double disk failure
and your DB file is dead as a dodo, at least eventually

If your cold storage were to be just as reliable as the rest, that concept
could work, but still lead to surprises, for example when booting a VM
after a long with most of its data having been pushed to the slow pool.

> Or may be that kind of data could be moved to a LTO tape library ...
> just thinking ! :)
> 
A dedicated pool (something with replica 1 or 2 and backed by RAID6 OSDs
for example or an EC pool) and "manual" moving would likely be better
suited than the approach above.

Christian

> Thanks for feedback and suggestion on how to handle data you "never will
> use again but you have to have them" :)
> 
> Regards . Götz
> 
> 
> 
> 


-- 
Christian BalzerNetwork/Systems Engineer
ch...@gol.com   Global OnLine Japan/Rakuten Communications
http://www.gol.com/
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com


[ceph-users] thoughts about Cache Tier Levels

2016-07-20 Thread Götz Reinicke - IT Koordinator
Hi,

currently there are two levels I know of: storage- and cachepool. From
our workload I do expect an third "level" of data, which will stay
currently in the storagepool as well.

Has anyone as we been thinking of data which could be moved even deeper
in that tiering, e.g. have SSD cache, fast lots of HDD SAS something
2-4TB size for storage pool and e.g. 8TB so whatever for an "archive" level?

From my pov that would be for data which are good and nice to have in
the cluster, but dont have to be as reliable and fast as the storage
tier data.

Or may be that kind of data could be moved to a LTO tape library ...
just thinking ! :)

Thanks for feedback and suggestion on how to handle data you "never will
use again but you have to have them" :)

Regards . Götz






smime.p7s
Description: S/MIME Cryptographic Signature
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com