Re: [ceph-users] BlueStore WAL or DB devices on a distant SSD ?

[David Turner]

Specifying them to be the same device is redundant and not necessary.  They
will be put on the bluestore device by default unless you specify them to
go to another device.

On Fri, Aug 18, 2017 at 9:17 AM Hervé Ballans 
wrote:

> Le 16/08/2017 à 16:19, David Turner a écrit :
>
> Would reads and writes to the SSD on another server be faster than reads
> and writes to HDD on the local server? If the answer is no, then even if
> this was possible it would be worse than just putting your WAL and DB on
> the same HDD locally.  I don't think this is a use case the devs planned
> for.
>
> You can definitely put multiple WAL land DB partitions on a single SSD.
>
>
> Thanks David for your reply.
> Indeed, I have no idea if I/O operations on a distant SSD are faster than
> on a local HDD ?..I thought naively that in the case of a 10Gbs network,
> the speed of the bandwidth was always higher than that of the disk,
> whatever its technology, but it's true that many other factors come into
> play (in any case, it's an interesting problem, if I have time, I will try
> to do some benchmarks...)
>
> In relation to your answer, I have another question please, in the case of
> an OSDs server composed only with HDD disks, what is the best practice :
> just configure each OSD as a single BlueStore storage device or add on each
> disk both WAL and DB devices (e.g. ceph-disk prepare --bluestore /dev/sdd
> --block.wal /dev/sdd --block.db /dev/sdd) ?
>
> I ask this because the online documentation on "BlueStore Config
> Reference" states that adding these devices are useful only if the device
> used is faster than the primary device ?
>
> Thanks again,
> Hervé
>
>
> On Wed, Aug 16, 2017, 6:04 AM Hervé Ballans 
> wrote:
>
>> Hi,
>>
>> We are currently running two Proxmox/ceph clusters that work perfectly
>> (since 2014) and thank to this succesful experience, we plan to install
>> a new Ceph cluster for storage of our computing cluster.
>>
>> Until now, we only used RBD (virtualization context) but now we want to
>> use CephFS for this new cluster (separated from the other two, hardware
>> is different and dedicated for this new clusters).
>>
>> I'm interested in testing a CephFS cluster with BlueStore as a backend
>> storage.
>>
>> I have several OSDs servers (with a dozen SATA HDDs on each) but some do
>> not have an additional SSD disk (only 3 of the servers have an
>> additional SSD).
>>
>> My question is about BlueStore WAL/DB devices. When I read the
>> documentation, it seems that adding both WAL and DB devices improve
>> BlueStore performances.
>>
>> But, can we configure these devices on a distant SSD (I mean on a SSD
>> which is not on the local OSDs server but on an another machine which is
>> on the same Ceph cluster) ?
>>
>> If yes, can I configure mulitple WAL or DB devices on the same SSD ?
>>
>> And finally, is it relevant to do that (I mean in term of performance) ?
>>
>> Hoping to have been clear on my context, thanks in advance for your
>> reply  or your reflection.
>>
>> Regards,
>>
>> Hervé
>>
>>
>> ___
>> ceph-users mailing list
>> ceph-users@lists.ceph.com
>> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
>>
>
> ___
> ceph-users mailing list
> ceph-users@lists.ceph.com
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
>
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

Re: [ceph-users] BlueStore WAL or DB devices on a distant SSD ?

[Hervé Ballans]

Le 16/08/2017 à 16:19, David Turner a écrit :


Would reads and writes to the SSD on another server be faster than 
reads and writes to HDD on the local server? If the answer is no, then 
even if this was possible it would be worse than just putting your WAL 
and DB on the same HDD locally.  I don't think this is a use case the 
devs planned for.


You can definitely put multiple WAL land DB partitions on a single SSD.



Thanks David for your reply.
Indeed, I have no idea if I/O operations on a distant SSD are faster 
than on a local HDD ?..I thought naively that in the case of a 10Gbs 
network, the speed of the bandwidth was always higher than that of the 
disk, whatever its technology, but it's true that many other factors 
come into play (in any case, it's an interesting problem, if I have 
time, I will try to do some benchmarks...)


In relation to your answer, I have another question please, in the case 
of an OSDs server composed only with HDD disks, what is the best 
practice : just configure each OSD as a single BlueStore storage device 
or add on each disk both WAL and DB devices (e.g. ceph-disk prepare 
--bluestore /dev/sdd --block.wal /dev/sdd --block.db /dev/sdd) ?


I ask this because the online documentation on "BlueStore Config 
Reference" states that adding these devices are useful only if the 
device used is faster than the primary device ?


Thanks again,
Hervé



On Wed, Aug 16, 2017, 6:04 AM Hervé Ballans 
> wrote:


Hi,

We are currently running two Proxmox/ceph clusters that work perfectly
(since 2014) and thank to this succesful experience, we plan to
install
a new Ceph cluster for storage of our computing cluster.

Until now, we only used RBD (virtualization context) but now we
want to
use CephFS for this new cluster (separated from the other two,
hardware
is different and dedicated for this new clusters).

I'm interested in testing a CephFS cluster with BlueStore as a backend
storage.

I have several OSDs servers (with a dozen SATA HDDs on each) but
some do
not have an additional SSD disk (only 3 of the servers have an
additional SSD).

My question is about BlueStore WAL/DB devices. When I read the
documentation, it seems that adding both WAL and DB devices improve
BlueStore performances.

But, can we configure these devices on a distant SSD (I mean on a SSD
which is not on the local OSDs server but on an another machine
which is
on the same Ceph cluster) ?

If yes, can I configure mulitple WAL or DB devices on the same SSD ?

And finally, is it relevant to do that (I mean in term of
performance) ?

Hoping to have been clear on my context, thanks in advance for your
reply  or your reflection.

Regards,

Hervé


___
ceph-users mailing list
ceph-users@lists.ceph.com 
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com



___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

Re: [ceph-users] BlueStore WAL or DB devices on a distant SSD ?

[David Turner]

Would reads and writes to the SSD on another server be faster than reads
and writes to HDD on the local server? If the answer is no, then even if
this was possible it would be worse than just putting your WAL and DB on
the same HDD locally.  I don't think this is a use case the devs planned
for.

You can definitely put multiple WAL land DB partitions on a single SSD.

On Wed, Aug 16, 2017, 6:04 AM Hervé Ballans 
wrote:

> Hi,
>
> We are currently running two Proxmox/ceph clusters that work perfectly
> (since 2014) and thank to this succesful experience, we plan to install
> a new Ceph cluster for storage of our computing cluster.
>
> Until now, we only used RBD (virtualization context) but now we want to
> use CephFS for this new cluster (separated from the other two, hardware
> is different and dedicated for this new clusters).
>
> I'm interested in testing a CephFS cluster with BlueStore as a backend
> storage.
>
> I have several OSDs servers (with a dozen SATA HDDs on each) but some do
> not have an additional SSD disk (only 3 of the servers have an
> additional SSD).
>
> My question is about BlueStore WAL/DB devices. When I read the
> documentation, it seems that adding both WAL and DB devices improve
> BlueStore performances.
>
> But, can we configure these devices on a distant SSD (I mean on a SSD
> which is not on the local OSDs server but on an another machine which is
> on the same Ceph cluster) ?
>
> If yes, can I configure mulitple WAL or DB devices on the same SSD ?
>
> And finally, is it relevant to do that (I mean in term of performance) ?
>
> Hoping to have been clear on my context, thanks in advance for your
> reply  or your reflection.
>
> Regards,
>
> Hervé
>
>
> ___
> ceph-users mailing list
> ceph-users@lists.ceph.com
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
>
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com