[lustre-discuss] Max Single OSS throughput is not crossing 7 GB/s Reads

[Andreas Dilger]

On Jun 28, 2022, at 21:51, Karan Singh via lustre-discuss <lustre-discuss at lists.lustre.org<mailto:lustre-discuss at lists.lustre.org>> wrote:

Hi team

Below are the details :
Using 40 lustre docker clients running on 4 x Dell R750 with each lustre docker client running the below mentioned command

${xfersize} = ${blksize}
${blksize} = 128M, 64M, 32M, 16M, 8M, 4M, 2M, 1M

mpirun -hostfile nodefile --map-by node -np 16 /usr/local/bin/ior -a POSIX --posix.odirect -v -i 50 -g -F -e -k -o /mnt/beegfs/test.ior

I think the problem is pretty clear right here - you are using BeeGFS for your testing, not Lustre...  :-)

-r -t ${xfersize} -b ${blksize} -O summaryFile=./${TESTDIR}/iorresult_SeqWrite_p${numthreads}_bs${blksize}_tf${xfersize}.json,summaryFormat=JSON


<image.png>

please let me know what is the bottleneck in the above setup ?
Also if you need more info , will be provide it right away .

Cheers, Andreas
--
Andreas Dilger
Lustre Principal Architect
Whamcloud







-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20220629/2163f583/attachment.htm>