massive rbd speed decrease with latest git master

massive rbd speed decrease with latest git master

[Stefan Priebe - Profihost AG]

Hello list,

while updating ceph to the latest master i noticed this:

Commit a01b112d71a0b6a1bb206d53867d13536d17bbf6:
rand. 4k WRITE: 3000 iop/s

Commit 72a710ac47355ae3ca1d055be1496c91956d549d:
rand. 4k WRITE: 900 iop/s

Same for rand 4k read - haven't tested sequential stuff.

Stefan

Re: massive rbd speed decrease with latest git master

[Stefan Priebe - Profihost AG]

Am 05.11.2012 11:39, schrieb Stefan Priebe - Profihost AG:
> Hello list,
>
> while updating ceph to the latest master i noticed this:
>
> Commit a01b112d71a0b6a1bb206d53867d13536d17bbf6:
> rand. 4k WRITE: 3000 iop/s
>
> Commit 72a710ac47355ae3ca1d055be1496c91956d549d:
> rand. 4k WRITE: 900 iop/s
>
> Same for rand 4k read - haven't tested sequential stuff.

Now i'm strangely seeing this slow performance with old commit too. 
Pretty strange. So please ignore my mail until i made further tests.

Stefan

Re: massive rbd speed decrease with latest git master

[Stefan Priebe]

Am 05.11.2012 11:39, schrieb Stefan Priebe - Profihost AG:
> Hello list,
>
> while updating ceph to the latest master i noticed this:
>
> Commit a01b112d71a0b6a1bb206d53867d13536d17bbf6:
> rand. 4k WRITE: 3000 iop/s
>
> Commit 72a710ac47355ae3ca1d055be1496c91956d549d:
> rand. 4k WRITE: 900 iop/s
>
> Same for rand 4k read - haven't tested sequential stuff.

OK i found out what the issue is. I get 900 iop/s when compiled with 
syncfs support and i get 3000-6000 iop/s when compiled without syncfs 
support.

Strange shouldn't syncfs improve the performance?

Note: ceph is running inside KVM right now for testing with an iscsi 
based backend storage. My real testhw will arrive tomorrow.

Greets,
Stefan

Re: massive rbd speed decrease with latest git master

[Mark Nelson]

On 11/05/2012 12:48 PM, Stefan Priebe wrote:
> Am 05.11.2012 11:39, schrieb Stefan Priebe - Profihost AG:
>> Hello list,
>>
>> while updating ceph to the latest master i noticed this:
>>
>> Commit a01b112d71a0b6a1bb206d53867d13536d17bbf6:
>> rand. 4k WRITE: 3000 iop/s
>>
>> Commit 72a710ac47355ae3ca1d055be1496c91956d549d:
>> rand. 4k WRITE: 900 iop/s
>>
>> Same for rand 4k read - haven't tested sequential stuff.
>
> OK i found out what the issue is. I get 900 iop/s when compiled with
> syncfs support and i get 3000-6000 iop/s when compiled without syncfs
> support.
>
> Strange shouldn't syncfs improve the performance?
>
> Note: ceph is running inside KVM right now for testing with an iscsi
> based backend storage. My real testhw will arrive tomorrow.

Very strange.  I'm curious if you continue to see this on your real test 
hardware.

>
> Greets,
> Stefan
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Mark

Re: massive rbd speed decrease with latest git master

[Stefan Priebe]

Am 05.11.2012 19:58, schrieb Mark Nelson:
> On 11/05/2012 12:48 PM, Stefan Priebe wrote:
>> Am 05.11.2012 11:39, schrieb Stefan Priebe - Profihost AG:
>>> Hello list,
>>>
>>> while updating ceph to the latest master i noticed this:
>>>
>>> Commit a01b112d71a0b6a1bb206d53867d13536d17bbf6:
>>> rand. 4k WRITE: 3000 iop/s
>>>
>>> Commit 72a710ac47355ae3ca1d055be1496c91956d549d:
>>> rand. 4k WRITE: 900 iop/s
>>>
>>> Same for rand 4k read - haven't tested sequential stuff.
>>
>> OK i found out what the issue is. I get 900 iop/s when compiled with
>> syncfs support and i get 3000-6000 iop/s when compiled without syncfs
>> support.
>>
>> Strange shouldn't syncfs improve the performance?
>>
>> Note: ceph is running inside KVM right now for testing with an iscsi
>> based backend storage. My real testhw will arrive tomorrow.
>
> Very strange.  I'm curious if you continue to see this on your real test
> hardware.

Yes - will report back tomorrow.

Stefan