FIXED: 11/5/2014 Weekly Ceph Performance Meeting IS ON!

FIXED: 11/5/2014 Weekly Ceph Performance Meeting IS ON!

[Mark Nelson]

Hi All,

Got the date wrong the first time.  11/5/2014 is correct!  8AM PST as 
usual.  We are going to try a little experiment and leave the Agenda 
blank for you guys to fill in. If you are planning on attending and have 
something you want to discuss, please add it!

We've also added a projects and backlog section at the top of the 
etherpad for on-going efforts and the folks that are working on them. 
For any projects or names of people we've missed, please update!

Here's the links:

Etherpad URL:
http://pad.ceph.com/p/performance_weekly

To join the Meeting:
https://bluejeans.com/268261044

To join via Browser:
https://bluejeans.com/268261044/browser

To join with Lync:
https://bluejeans.com/268261044/lync


To join via Room System:
Video Conferencing System: bjn.vc -or- 199.48.152.152
Meeting ID: 268261044

To join via Phone:
1) Dial:
           +1 408 740 7256
           +1 888 240 2560(US Toll Free)
           +1 408 317 9253(Alternate Number)
           (see all numbers - http://bluejeans.com/numbers)
2) Enter Conference ID: 268261044

Mark

Re: FIXED: 11/5/2014 Weekly Ceph Performance Meeting IS ON!

[Andreas Bluemle]

[-- Attachment #1: Type: text/plain, Size: 2108 bytes --]

Hi,

I have spent some time on using LTTng tracepoints
in order to determine the contributions to the latency for
individual write requests.

I have attached a table which shows the time spent in various
functions of the ceph-osd daemon for a write request to the
primary osd; I was using unmodified ceph-v0.86.

On the client side, I was using fio on rbd with
 - libaio and iodepth of 16
 - random write
 - 4 kByte requests

Maybe this is of interest.


Regards

Andreas



On Tue, 04 Nov 2014 09:49:15 -0600
Mark Nelson <mark.nelson@inktank.com> wrote:

> Hi All,
> 
> Got the date wrong the first time.  11/5/2014 is correct!  8AM PST as 
> usual.  We are going to try a little experiment and leave the Agenda 
> blank for you guys to fill in. If you are planning on attending and
> have something you want to discuss, please add it!
> 
> We've also added a projects and backlog section at the top of the 
> etherpad for on-going efforts and the folks that are working on them. 
> For any projects or names of people we've missed, please update!
> 
> Here's the links:
> 
> Etherpad URL:
> http://pad.ceph.com/p/performance_weekly
> 
> To join the Meeting:
> https://bluejeans.com/268261044
> 
> To join via Browser:
> https://bluejeans.com/268261044/browser
> 
> To join with Lync:
> https://bluejeans.com/268261044/lync
> 
> 
> To join via Room System:
> Video Conferencing System: bjn.vc -or- 199.48.152.152
> Meeting ID: 268261044
> 
> To join via Phone:
> 1) Dial:
>            +1 408 740 7256
>            +1 888 240 2560(US Toll Free)
>            +1 408 317 9253(Alternate Number)
>            (see all numbers - http://bluejeans.com/numbers)
> 2) Enter Conference ID: 268261044
> 
> Mark
> --
> 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
> 
> 



-- 
Andreas Bluemle                     mailto:Andreas.Bluemle@itxperts.de
Heinrich Boell Strasse 88           Phone: (+49) 89 4317582
D-81829 Muenchen (Germany)          Mobil: (+49) 177 522 0151

[-- Attachment #2: timestamps-primary-write.pdf --]
[-- Type: application/pdf, Size: 27438 bytes --]

Re: FIXED: 11/5/2014 Weekly Ceph Performance Meeting IS ON!

[Alexandre DERUMIER]

Hi Mark,

I'm not available for today (Next week I'll finally have a microphone ;)

But Here my progress about client side performance


qemu scaling
------------

I'm finally able to scale inside 1 guest vm with iothreads + krbd + virtio disks
I can reach now 90000iops in 1 guest vm (30000 iops by iothread)


-object iothread,id=iothread0 -object iothread,id=iothread1 -object iothread,id=iothread2
-device virtio-blk-pci,....,iothread=iothread0 -drive file=/dev/rbd/test/test
-device virtio-blk-pci,....,iothread=iothread1 -drive file=/dev/rbd/test/test
-device virtio-blk-pci,....,iothread=iothread2 -drive file=/dev/rbd/test/test

Currently It's not possible to bind multiple iothread on 1 virtio disk,
but you can define multiple virtio disk with same rbd volume,
and do some multipathing inside the guest vm.

I think qemu plan to implement multiqueue + multi iothread for same disk next year


krbd vs lirbd cpu usage
------------------------
Not to much progress,

but fio-librbd or qemu-librbd, I see around 3-4x more cpu usage with librbd.
That's really a bottleneck in qemu, I can't reach more than 25000iops with same iothread config.

I have some perf reports for analysis if someone would help.



Regards,

Alexandre




----- Mail original ----- 

De: "Mark Nelson" <mark.nelson@inktank.com> 
À: ceph-devel@vger.kernel.org 
Envoyé: Mardi 4 Novembre 2014 16:49:15 
Objet: FIXED: 11/5/2014 Weekly Ceph Performance Meeting IS ON! 

Hi All, 

Got the date wrong the first time. 11/5/2014 is correct! 8AM PST as 
usual. We are going to try a little experiment and leave the Agenda 
blank for you guys to fill in. If you are planning on attending and have 
something you want to discuss, please add it! 

We've also added a projects and backlog section at the top of the 
etherpad for on-going efforts and the folks that are working on them. 
For any projects or names of people we've missed, please update! 

Here's the links: 

Etherpad URL: 
http://pad.ceph.com/p/performance_weekly 

To join the Meeting: 
https://bluejeans.com/268261044 

To join via Browser: 
https://bluejeans.com/268261044/browser 

To join with Lync: 
https://bluejeans.com/268261044/lync 


To join via Room System: 
Video Conferencing System: bjn.vc -or- 199.48.152.152 
Meeting ID: 268261044 

To join via Phone: 
1) Dial: 
+1 408 740 7256 
+1 888 240 2560(US Toll Free) 
+1 408 317 9253(Alternate Number) 
(see all numbers - http://bluejeans.com/numbers) 
2) Enter Conference ID: 268261044 

Mark 
-- 
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 
--
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