<font size=2 face="sans-serif">Just to clarify ( from man exports):</font><br><font size=2 face="sans-serif">" async This option
allows the NFS server to violate the NFS protocol and reply to requests
before any changes made by that request have been committed to stable
storage (e.g.</font><br><font size=2 face="sans-serif">
disc drive).</font><br><br><font size=2 face="sans-serif">
Using this option usually improves performance, but at the cost
that an unclean server restart (i.e. a crash) can cause data to be lost
or corrupted."</font><br><br><font size=2 face="sans-serif">With the Ganesha implementation in Spectrum
Scale, it was decided not to allow this violation - so this async export
options wasn't exposed.</font><br><font size=2 face="sans-serif">I believe that for those customers that
agree to take the risk, using async mount option ( from the client) will
achieve similar behavior.</font><br><font size=2 face="sans-serif"><br>Regards,<br><br>Tomer Perry<br>Scalable I/O Development (Spectrum Scale)<br>email: tomp@il.ibm.com<br>1 Azrieli Center, Tel Aviv 67021, Israel<br>Global Tel: +1 720 3422758<br>Israel Tel: +972 3 9188625<br>Mobile: +972 52 2554625<br></font><br><br><br><br><font size=1 color=#5f5f5f face="sans-serif">From:
</font><font size=1 face="sans-serif">"Olaf Weiser"
<olaf.weiser@de.ibm.com></font><br><font size=1 color=#5f5f5f face="sans-serif">To:
</font><font size=1 face="sans-serif">gpfsug main discussion
list <gpfsug-discuss@spectrumscale.org></font><br><font size=1 color=#5f5f5f face="sans-serif">Date:
</font><font size=1 face="sans-serif">17/10/2018 16:16</font><br><font size=1 color=#5f5f5f face="sans-serif">Subject:
</font><font size=1 face="sans-serif">Re: [gpfsug-discuss]
Preliminary conclusion: single client, single thread, small files - native
Scale vs NFS</font><br><font size=1 color=#5f5f5f face="sans-serif">Sent by:
</font><font size=1 face="sans-serif">gpfsug-discuss-bounces@spectrumscale.org</font><br><hr noshade><br><br><br><font size=2 face="sans-serif">Jallo Jan, <br>you can expect to get slightly improved numbers from the lower response
times of the HAWC ... but the loss of performance comes from the fact,
that <br>GPFS or (async kNFS) writes with multiple parallel threads - in opposite
to e.g. tar via GaneshaNFS comes with single threads fsync on each
file.. </font><font size=3><br></font><font size=2 face="sans-serif"><br>you'll never outperform e.g. 128 (maybe slower), but, parallel threads
(running write-behind) <---> with one single but fast
threads, ....</font><font size=3><br></font><font size=2 face="sans-serif"><br>so as Alex suggest.. if possible.. take gpfs client of kNFS for those
types of workloads..</font><font size=3><br><br><br></font><br><font size=3><br><br><br><br><br><br></font><font size=1 color=#5f5f5f face="sans-serif"><br>From: </font><font size=1 face="sans-serif">Jan-Frode
Myklebust <janfrode@tanso.net></font><font size=1 color=#5f5f5f face="sans-serif"><br>To: </font><font size=1 face="sans-serif">gpfsug
main discussion list <gpfsug-discuss@spectrumscale.org></font><font size=1 color=#5f5f5f face="sans-serif"><br>Date: </font><font size=1 face="sans-serif">10/17/2018
02:24 PM</font><font size=1 color=#5f5f5f face="sans-serif"><br>Subject: </font><font size=1 face="sans-serif">Re:
[gpfsug-discuss] Preliminary conclusion: single client, single thread,
small files - native Scale vs NFS</font><font size=1 color=#5f5f5f face="sans-serif"><br>Sent by: </font><font size=1 face="sans-serif">gpfsug-discuss-bounces@spectrumscale.org</font><font size=3><br></font><hr noshade><font size=3><br><br><br>Do you know if the slow throughput is caused by the network/nfs-protocol
layer, or does it help to use faster storage (ssd)? If on storage, have
you considered if HAWC can help?<br><br>I’m thinking about adding an SSD pool as a first tier to hold the active
dataset for a similar setup, but that’s mainly to solve the small file
read workload (i.e. random I/O ).<br><br><br>-jf<br>ons. 17. okt. 2018 kl. 07:47 skrev Alexander Saupp <</font><a href="mailto:Alexander.Saupp@de.ibm.com"><font size=3 color=blue><u>Alexander.Saupp@de.ibm.com</u></font></a><font size=3>>:</font><font size=2><br>Dear Mailing List readers,<br><br>I've come to a preliminary conclusion that explains the behavior in an
appropriate manner, so I'm trying to summarize my current thinking with
this audience.<b><br><br>Problem statement: </b><br>Big performance derivation between native GPFS (fast) and loopback NFS
mount on the same node (way slower) for single client, single thread, small
files workload.</font><font size=3><br></font><font size=2><b><br><br>Current explanation:</b><br>tar seems to use close() on files, not fclose(). That is an application
choice and common behavior. The ideas is to allow OS write caching to speed
up process run time.<br><br>When running locally on ext3 / xfs / GPFS / .. that allows async destaging
of data down to disk, somewhat compromising data for better performance.
<br>As we're talking about write caching on the same node that the application
runs on - a crash is missfortune but in the same failure domain.<br>E.g. if you run a compile job that includes extraction of a tar and the
node crashes you'll have to restart the entire job, anyhow.<br><br>The NFSv2 spec defined that NFS io's are to be 'sync', probably because
the compile job on the nfs client would survive if the NFS Server crashes,
so the failure domain would be different<br><br>NFSv3 in rfc1813 below acknowledged the performance impact and introduced
the 'async' flag for NFS, which would handle IO's similar to local IOs,
allowing to destage in the background.<br><br>Keep in mind - applications, independent if running locally or via NFS
can always decided to use the fclose() option, which will ensure that data
is destaged to persistent storage right away.<br>But its an applications choice if that's really mandatory or whether performance
has higher priority.<br><br>The linux 'sync' (man sync) tool allows to sync 'dirty' memory cache down
to disk - very filesystem independent.</font><font size=3><br></font><font size=2><br>-> single client, single thread, small files workload on GPFS can be
destaged async, allowing to hide latency and parallelizing disk IOs.<br>-> NFS client IO's are sync, so the second IO can only be started after
the first one hit non volatile memory -> much higher latency</font><font size=3><br></font><font size=2><br><br>The Spectrum Scale NFS implementation (based on ganesha) does not support
the async mount option, which is a bit of a pitty. There might also be
implementation differences compared to kernel-nfs, I did not investigate
into that direction.<br><br>However, the principles of the difference are explained for my by the above
behavior. <br><br>One workaround that I saw working well for multiple customers was to replace
the NFS client by a Spectrum Scale nsd client.<br>That has two advantages, but is certainly not suitable in all cases:<br>- Improved speed by efficent NSD protocol and NSD client side write caching<br>- Write Caching in the same failure domain as the application (on NSD client)
which seems to be more reasonable compared to NFS Server side write caching.</font><font size=3><br></font><font size=2><b><br>References:</b><br><br>NFS sync vs async</font><font size=3 color=blue><u><br></u></font><a href="https://tools.ietf.org/html/rfc1813" target="_blank"><font size=2 color=blue><u>https://tools.ietf.org/html/rfc1813</u></font></a><font size=2><i><br>The write throughput bottleneck caused by the synchronous definition of
write in the NFS version 2 protocol has been addressed by adding support
so that the NFS server can do unsafe writes.</i><br>Unsafe writes are writes which have not been committed to stable storage
before the operation returns. This specification defines a method for committing
these unsafe writes to stable storage in a reliable way.</font><font size=3><br></font><font size=2><b><br><br>sync() vs fsync()</b></font><font size=3 color=blue><u><br></u></font><a href="https://www.ibm.com/support/knowledgecenter/en/ssw_aix_72/com.ibm.aix.performance/using_sync_fsync_calls.htm" target="_blank"><font size=2 color=blue><u>https://www.ibm.com/support/knowledgecenter/en/ssw_aix_72/com.ibm.aix.performance/using_sync_fsync_calls.htm</u></font></a><font size=2><br>- An application program makes an fsync() call for a specified file. This
causes all of the pages that contain modified data for that file to be
written to disk. The writing is complete when the fsync() call returns
to the program.<br><br>- An application program makes a sync() call. This causes all of the file
pages in memory that contain modified data to be scheduled for writing
to disk. The writing is not necessarily complete when the sync() call returns
to the program.<br><br>- A user can enter the sync command, which in turn issues a sync() call.
Again, some of the writes may not be complete when the user is prompted
for input (or the next command in a shell script is processed).</font><font size=3><br></font><font size=2><b><br><br>close() vs fclose()</b><br>A successful close does not guarantee that the data has been successfully
saved to disk, as the kernel defers writes. It is not common for a file
system to flush the buffers when the stream is closed. If you need to be
sure that the data is<br>physically stored use fsync(2). (It will depend on the disk hardware at
this point.)</font><font size=3><br></font><font size=2><br><br>Mit freundlichen Grüßen / Kind regards</font><font size=3><b><br><br>Alexander Saupp</b></font><font size=3 color=#0060a0><br><br>IBM Systems, Storage Platform, EMEA Storage Competence Center</font><table width=800 style="border-collapse:collapse;"><tr height=8><td width=798 colspan=4 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><hr><tr height=8><td width=798 colspan=4 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><tr height=8><td width=87 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>Phone:</font><td width=319 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>+49
7034-643-1512</font><td width=247 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>IBM
Deutschland GmbH</font><td width=138 rowspan=4 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><tr height=8><td width=87 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>Mobile:</font><td width=319 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>+49-172
7251072</font><td width=247 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>Am
Weiher 24</font><tr height=8><td width=87 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>Email:</font><td width=319 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><a href="mailto:alexander.saupp@de.ibm.com" target="_blank"><font size=1 color=blue><u>alexander.saupp@de.ibm.com</u></font></a><td width=247 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>65451
Kelsterbach</font><tr height=8><td width=87 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><td width=319 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><td width=247 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#0060a0>Germany</font><tr height=8><td width=798 colspan=4 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><hr><tr height=8><td width=798 colspan=4 style="border-style:none none none none;border-color:#000000;border-width:0px 0px 0px 0px;padding:1px 1px;"><font size=1 color=#a2a2a2>IBM
Deutschland GmbH / Vorsitzender des Aufsichtsrats: Martin Jetter<br>Geschäftsführung: Matthias Hartmann (Vorsitzender), Norbert Janzen, Stefan
Lutz, Nicole Reimer, Dr. Klaus Seifert, Wolfgang Wendt<br>Sitz der Gesellschaft: Ehningen / Registergericht: Amtsgericht Stuttgart,
HRB 14562 / WEEE-Reg.-Nr. DE 99369940 </font></table><br><font size=3><br><br><br>_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at </font><a href="http://spectrumscale.org" target="_blank"><font size=3 color=blue><u>spectrumscale.org</u></font></a><font size=3 color=blue><u><br></u></font><a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss" target="_blank"><font size=3 color=blue><u>http://gpfsug.org/mailman/listinfo/gpfsug-discuss</u></font></a><font size=3 color=blue><u>[attachment
"ecblank.gif" deleted by Olaf Weiser/Germany/IBM] [attachment
"19995626.gif" deleted by Olaf Weiser/Germany/IBM] [attachment
"ecblank.gif" deleted by Olaf Weiser/Germany/IBM] </u></font><tt><font size=2>_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at spectrumscale.org</font></tt><font size=3 color=blue><u><br></u></font><a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss"><tt><font size=2 color=blue><u>http://gpfsug.org/mailman/listinfo/gpfsug-discuss</u></font></tt></a><font size=3><br><br></font><br><tt><font size=2>_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at spectrumscale.org<br></font></tt><a href="http://gpfsug.org/mailman/listinfo/gpfsug-discuss"><tt><font size=2>http://gpfsug.org/mailman/listinfo/gpfsug-discuss</font></tt></a><tt><font size=2><br></font></tt><br><br><BR>