<html><body><p><tt><font size="2">> "It has a significant performance penalty for small files in large <br>> block size filesystems"</font></tt><br><br><font size="2">Aaron,</font><br><br><font size="2">Below are mdtest results for a test we ran for CORAL - file size was 32k.</font><br><br><font size="2">We have not gone back and ran the test on a file system formatted without > 32 subblocks. We'll do that at some point...</font><br><br><font size="2">-Bill</font><br><br><font size="2" face="Courier New">-- started at 10/28/2017 17:51:38 --</font><br><br><font size="2" face="Courier New">mdtest-1.9.3 was launched with 228 total task(s) on 12 node(s)</font><br><font size="2" face="Courier New">Command line used: /tmp/mdtest-binary-dir/mdtest -d /ibm/fs2-16m-10/mdtest-60000 -i 3 -n 294912 -w 32768 -C -F -r -p 360 -u -y</font><br><font size="2" face="Courier New">Path: /ibm/fs2-16m-10</font><br><font size="2" face="Courier New">FS: 128.1 TiB Used FS: 0.3% Inodes: 476.8 Mi Used Inodes: 0.0%</font><br><br><font size="2" face="Courier New">228 tasks, 67239936 files</font><br><br><font size="2" face="Courier New">SUMMARY: (of 3 iterations)</font><br><font size="2" face="Courier New"> Operation Max Min Mean Std Dev</font><br><font size="2" face="Courier New"> --------- --- --- ---- -------</font><br><font size="2" face="Courier New"> File creation : 51953.498 50558.517 51423.221 616.643</font><br><font size="2" face="Courier New"> File stat : 0.000 0.000 0.000 0.000</font><br><font size="2" face="Courier New"> File read : 0.000 0.000 0.000 0.000</font><br><font size="2" face="Courier New"> File removal : 96746.376 92149.535 94658.774 1900.187</font><br><font size="2" face="Courier New"> Tree creation : 1.588 0.070 0.599 0.700</font><br><font size="2" face="Courier New"> Tree removal : 0.213 0.034 0.097 0.082</font><br><br><font size="2" face="Courier New">-- finished at 10/28/2017 19:51:54 --</font><br><font size="2"><br>Bill Hartner<br>IBM Systems<br>Scalable I/O Development<br>Austin, Texas<br>bhartner@us.ibm.com<br>home office 512-784-0980<br></font><br><br><tt><font size="2">gpfsug-discuss-bounces@spectrumscale.org wrote on 11/29/2017 04:41:48 PM:<br><br>> From: Aaron Knister <aaron.knister@gmail.com></font></tt><br><tt><font size="2">> To: gpfsug main discussion list <gpfsug-discuss@spectrumscale.org></font></tt><br><tt><font size="2">> Date: 11/29/2017 04:42 PM</font></tt><br><tt><font size="2">> Subject: Re: [gpfsug-discuss] Online data migration tool</font></tt><br><tt><font size="2">> Sent by: gpfsug-discuss-bounces@spectrumscale.org</font></tt><br><tt><font size="2">> <br>> Thanks, Nikhil. Most of that was consistent with my understnading, <br>> however I was under the impression that the >32 subblocks code is <br>> required to achieve the touted 50k file creates/second that Sven has<br>> talked about a bunch of times:<br>> <br>> <a href="http://files.gpfsug.org/presentations/2017/Manchester/08_Research_Topics.pdf">http://files.gpfsug.org/presentations/2017/Manchester/08_Research_Topics.pdf</a><br>> <a href="http://files.gpfsug.org/presentations/2017/Ehningen/31_-_SSUG17DE_-">http://files.gpfsug.org/presentations/2017/Ehningen/31_-_SSUG17DE_-</a><br>> _Sven_Oehme_-_News_from_Research.pdf<br>> <a href="http://files.gpfsug.org/presentations/2016/SC16/12_-">http://files.gpfsug.org/presentations/2016/SC16/12_-</a><br>> _Sven_Oehme_Dean_Hildebrand_-_News_from_IBM_Research.pdf<br></font></tt><br><tt><font size="2">> from those presentations regarding 32 subblocks:<br>> <br>> "It has a significant performance penalty for small files in large <br>> block size filesystems"<br></font></tt><br><tt><font size="2">> although I'm not clear on the specific definition of "large". Many <br>> filesystems I encounter only have a 1M block size so it may not <br>> matter there, although that same presentation clearly shows the <br>> benefit of larger block sizes which is yet *another* thing for which<br>> a migration tool would be helpful.<br></font></tt><br><tt><font size="2">> -Aaron</font></tt><br><tt><font size="2">> <br>> On Wed, Nov 29, 2017 at 2:08 PM, Nikhil Khandelwal <nikhilk@us.ibm.com> wrote:</font></tt><br><tt><font size="2">> Hi,<br>> <br>> I would like to clarify migration path to 5.0.0 from 4.X.X clusters.<br>> For all Spectrum Scale clusters that are currently at 4.X.X, it is <br>> possible to migrate to 5.0.0 with no offline data migration and no <br>> need to move data. Once these clusters are at 5.0.0, they will <br>> benefit from the performance improvements, new features (such as <br>> file audit logging), and various enhancements that are included in 5.0.0.<br>> <br>> That being said, there is one enhancement that will not be applied <br>> to these clusters, and that is the increased number of sub-blocks <br>> per block for small file allocation. This means that for file <br>> systems with a large block size and a lot of small files, the <br>> overall space utilization will be the same it currently is in 4.X.X.<br>> Since file systems created at 4.X.X and earlier used a block size <br>> that kept this allocation in mind, there should be very little <br>> impact on existing file systems.<br>> <br>> Outside of that one particular function, the remainder of the <br>> performance improvements, metadata improvements, updated <br>> compatibility, new functionality, and all of the other enhancements <br>> will be immediately available to you once you complete the upgrade <br>> to 5.0.0 -- with no need to reformat, move data, or take your data offline.<br>> <br>> I hope that clarifies things a little and makes the upgrade path <br>> more accessible.<br>> <br>> Please let me know if there are any other questions or concerns.<br>> <br>> Thank you,<br>> Nikhil Khandelwal<br>> Spectrum Scale Development<br>> Client Adoption</font></tt><br><tt><font size="2">> <br>> _______________________________________________<br>> gpfsug-discuss mailing list<br>> gpfsug-discuss at spectrumscale.org<br>> <a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss">http://gpfsug.org/mailman/listinfo/gpfsug-discuss</a><br></font></tt><br><tt><font size="2">> _______________________________________________<br>> gpfsug-discuss mailing list<br>> gpfsug-discuss at spectrumscale.org<br>> <a href="https://urldefense.proofpoint.com/v2/url?">https://urldefense.proofpoint.com/v2/url?</a><br>> u=http-3A__gpfsug.org_mailman_listinfo_gpfsug-2Ddiscuss&d=DwICAg&c=jf_iaSHvJObTbx-<br>> siA1ZOg&r=Ew59QH6nxuyx6oTs7a8AYX7kKG3gaWUGDGo5ZZr3wQ4&m=DHoqgBeMFgcM0LpXEI0VCYvvb8ollct5aSYUDln2t68&s=iOxGm-853L_W0XkB3jGsGzCTVlSYUvANOTSewcR_Ue8&e=<br></font></tt><BR>
</body></html>