Fwd: Re: performance issue with jewel on ubuntu xenial (kernel)

[Yoann Moulin <yoann.moulin@epfl.ch>]

Hello,

just forward this here, there is a big performance drop with kernel 4.4 compare
to 4.2 on Jewel.

Does anyone here have seen that before ?

Best Regards,

Yoann Moulin

-------- Message transféré --------
Sujet : Re: [ceph-users] performance issue with jewel on ubuntu xenial (kernel)
Date : Fri, 1 Jul 2016 16:37:41 +0200
De : Yoann Moulin <yoann.moulin@epfl.ch>

Hello,

>>>>>>> I found a performance drop between kernel 3.13.0-88 (default kernel on Ubuntu
>>>>>>> Trusty 14.04) and kernel 4.4.0.24.14 (default kernel on Ubuntu Xenial 16.04)
>>>>>>>
>>>>>>> ceph version is Jewel (10.2.2).
>>>>>>> All tests have been done under Ubuntu 14.04
>>>>>>
>>>>>> Knowing that you also have an internalis cluster on almost identical
>>>>>> hardware, can you please let the list know whether you see the same
>>>>>> behavior (severely reduced throughput on a 4.4 kernel, vs. 3.13) on
>>>>>> that cluster as well?
>>>>>
>>>>> ceph version is infernalis (9.2.0)
>>>>>
>>>>> Ceph osd Benchmark:
>>>>>
>>>>> Kernel 3.13.0-88-generic : ceph tell osd.ID => average ~84MB/s
>>>>> Kernel 4.2.0-38-generic  : ceph tell osd.ID => average ~90MB/s
>>>>> Kernel 4.4.0-24-generic  : ceph tell osd.ID => average ~75MB/s
>>>>>
>>>>> The slow down is not as much as I have with Jewel but it is still present.
>>>>
>>>> But this is not on precisely identical hardware, is it?
>>>
>>> All the benchmarks were run on strictly identical hardware setups per node.
>>> Clusters differ slightly in sizes (infernalis vs jewel) but nodes and OSDs are identical.
>>
>> One thing differ in the osd configuration, on the Jewel cluster, we have journal
>> on disk, on the Infernalis cluster, we have journal on SSD (S3500)
>>
>> I can restart my test on a Jewel cluster with journal on SSD if needed.
>> I can do as well a test on an Infernalis cluster with journal on disk.
> 
> I'd suggest that the second option is probably more meaningful to test.

I did new benchmarks on 3 clusters. Each cluster has 3 nodes strictly identical.
Each node has 10 OSDs. Journals are on the disk.

bench5 : Ubuntu 14.04 / Ceph Infernalis
bench6 : Ubuntu 14.04 / Ceph Jewel
bench7 : Ubuntu 16.04 / Ceph jewel

this is the average of 2 runs of "ceph tell osd.* bench" on each cluster (2 x 30
OSDs)

bench5 / 14.04 / Infernalis / kernel 3.13 :  54.35 MB/s
bench6 / 14.04 / Jewel      / kernel 3.13 :  86.47 MB/s

bench5 / 14.04 / Infernalis / kernel 4.2  :  63.38 MB/s
bench6 / 14.04 / Jewel      / kernel 4.2  : 107.75 MB/s
bench7 / 16.04 / Jewel      / kernel 4.2  : 101.54 MB/s

bench5 / 14.04 / Infernalis / kernel 4.4  :  53.61 MB/s
bench6 / 14.04 / Jewel      / kernel 4.4  :  65.82 MB/s
bench7 / 16.04 / Jewel      / kernel 4.4  :  61.57 MB/s

If needed, I have the raw output of "ceph tell osd.* bench"

> What I find curious is that no-one else on the list has apparently run
> into this. Any Ubuntu xenial users out there, or perhaps folks on
> trusty who choose to install linux-image-generic-lts-xenial?

Anyone to try on their side if they have the same behaviour ?

Cheers,

-- 
Yoann Moulin
EPFL IC-IT

--
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