10/22/2014 Weekly Ceph Performance Meeting

Discussion:

Mark Nelson

2014-10-22 14:21:55 UTC

Hi All,

Meeting is on at 8AM PST as usual! 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 ***@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Sage Weil

2014-10-22 16:24:50 UTC

Permalink

Hi Everyone,

Just a reminder that there won't be a performance call next week because
of CDS, which is Tuesday and Wednesday. The tentative schedule is here:

https://wiki.ceph.com/Planning/CDS/Hammer_(Oct_2014)

If there are key performance items we should be focusing on that aren't
yet captured by a blueprint and session or that would be useful to discuss
with a broader audience during CDS, speak up soon so we can get a slot
added to the scheduled.

I already mentioned to Matt that it would be good to talk about this
encode/decode work, and I believe Haomai has a proposal there as well.

sage

Post by Mark Nelson
Hi All,
http://pad.ceph.com/p/performance_weekly
https://bluejeans.com/268261044
https://bluejeans.com/268261044/browser
https://bluejeans.com/268261044/lync
Video Conferencing System: bjn.vc -or- 199.48.152.152
Meeting ID: 268261044
+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
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 ***@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Haomai Wang

2014-10-23 10:00:41 UTC

Permalink

https://wiki.ceph.com/Planning/Blueprints/Submissions/Fixed_memory_layout_for_Message%2F%2FOp_passing

@sage, thanks

Post by Sage Weil
Hi Everyone,
Just a reminder that there won't be a performance call next week because
https://wiki.ceph.com/Planning/CDS/Hammer_(Oct_2014)
If there are key performance items we should be focusing on that aren't
yet captured by a blueprint and session or that would be useful to discuss
with a broader audience during CDS, speak up soon so we can get a slot
added to the scheduled.
I already mentioned to Matt that it would be good to talk about this
encode/decode work, and I believe Haomai has a proposal there as well.
sage

--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
More majordomo info at http://vger.kernel.org/majordomo-info.html

--
Best Regards,

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

Zhang, Jian

2014-10-23 07:37:01 UTC

Permalink

Hi all,
Regarding on the cache tiering performance topic discussed on yesterday's meeting:
- Need to test performance with commit to avoid promotion on first read:
https://github.com/ceph/ceph/commit/79d1aff1821bc9f21477636df4d0d4e57f2cd008
Zhiqiang actually did some tests based on https://github.com/ceph/ceph/commit/0ed3adc1e0a74bf9548d1d956aece11f019afee0 .
We actually observed with Promote on 2nd read, we can get up to 2x IOPS improvement, (although still far below expectation), and 66% improvement on latency on the following test configuration.

Test configuration:
* Hardware:
2 OSD nodes with 10G Nic, each with 8x 1T HDD and 6x 400G DC3700 SSD, 2 SSDs used for Journal, while 4 SSD used for cache tiering.
* Cache Tiering Setting:
Cache size: 1000G (target_max_bytes: 1000000000000); Dirty ratio: 20% (cache_target_dirty_ratio: 0.2); Full ratio: 20% (cache_target_full_ratio: 0.2)
* Benchmark:
Fio 4K random Read/write/Mixed tests.
* Results:
Without cache tiering, the performance for 4K random read is: 1021, with latency ~3000ms (QD=256).
With cache tiering, the performance for 4K random read is: 359, with latency ~6000ms.
With cache tiering + promote on second read, 4K random read performance is: ~800, with latency ~1500ms.

We can share the data either during CDS https://wiki.ceph.com/Planning/Blueprints/Hammer/osd%3A_tiering%3A_fine-grained_promotion_unit or next weekly performance meeting.

Thanks
Jian

-----Original Message-----
From: ceph-devel-***@vger.kernel.org [mailto:ceph-devel-***@vger.kernel.org] On Behalf Of Mark Nelson
Sent: Wednesday, October 22, 2014 10:22 PM
To: ceph-***@vger.kernel.org
Subject: 10/22/2014 Weekly Ceph Performance Meeting

Hi All,

Meeting is on at 8AM PST as usual! 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 ***@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 ***@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html