Hi Rene,
Salmon, Rene wrote:
Hi list,
Just trying to benchmark our storage array to see how well it would
perform with Lustre. Looked through the manual and the Readme that came
with the Lustre I/O toolkit and just have some questions.
I am doing all this on a single box that is acting as an MDS,OSS, and
client that has the storage array attached to it. I simply want to
measure how the disk array will perform with lustre so for now I am not
using any of the network stuff.
The only option I am passing to obfilter-survey is "case=disk".
Following are the write results. Can some explain why the min part of
the [min max] field define as:
[min max] Minimum and maximum instantaneous bandwidths seen on any
individual OST.
To obtain Min and Max values It first finds minimum/maximum bandwidth
from each individual
OST taken at regular instances say one second. From those bandwidths it
finds Min and Max among all OST's
if mostly zero? I did not have that happen when I ran sgpdd_survey.
ost 1 sz 16777216K rsz 1024 obj 1 thr 1 write 260.68 [ 236.76,
267.57]
ost 1 sz 16777216K rsz 1024 obj 1 thr 2 write 285.99 [ 234.79,
289.99]
ost 1 sz 16777216K rsz 1024 obj 1 thr 4 write 284.02 [ 0.00,
323.40]
ost 1 sz 16777216K rsz 1024 obj 1 thr 8 write 109.94 [ 0.00,
319.77]
ost 1 sz 16777216K rsz 1024 obj 2 thr 2 write 182.90 [ 0.00,
265.21]
ost 1 sz 16777216K rsz 1024 obj 2 thr 4 write 113.07 [ 0.00,
255.98]
ost 1 sz 16777216K rsz 1024 obj 2 thr 8 write 81.01 [ 0.00,
263.63]
ost 1 sz 16777216K rsz 1024 obj 4 thr 4 write 95.06 [ 0.00,
234.26]
ost 1 sz 16777216K rsz 1024 obj 4 thr 8 write 58.80 [ 0.00,
213.68]
ost 1 sz 16777216K rsz 1024 obj 8 thr 8 write 47.20 [ 0.00,
186.81]
Can you please send '.detail' file for the test you ran? by default it
is available at /tmp/
-Jitendra
_______________________________________________
Lustre-discuss mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
