Hi Pinkesh,
The dual runs of the IO500 benchmark for this list are intentional,
and documented in the README-ISC20.txt file in the source tree.
This is to allow comparison between the historical io500.sh script and the
new C application that runs the same IOR, mdtest, and find commands.
Please submit both results for ISC'20. 

We wanted to be sure that the transition to the new C-app didn't
introduce any errors in the results. The need to run the benchmark
twice will hopefully be gone for the SC'20 list. 

Cheers, Andreas(*)
* speaking on my own behalf and not on behalf of the IO500 board

On May 29, 2020, at 16:02, Pinkesh Valdria via IO-500 <io-500@vi4io.org> wrote:



Hello IO-500 experts, 

 

I am trying to configure io500 .  When I run it,  it runs twice,  first one is regular and 2nd one is called “Running the C version of the benchmark now”.   Is it because I misconfigured it or is it required to run both, starting 2020 ?     My config*.ini file is below. 

 

 

[root@inst-q7cdd-good-crow io500-app]# ./io500.sh config-test1.ini

System:  inst-q7cdd-good-crow

…..

Running the IO500 Benchmark now

[Creating] directories

…..

[Summary] Results files in ./results//2020.05.29-17.31.34-scr

[Summary] Data files in ./out//2020.05.29-17.31.34-scr

[RESULT] BW   phase 1            ior_easy_write                6.188 GiB/s : time 357.44 seconds

[RESULT] BW   phase 2            ior_hard_write                1.132 GiB/s : time 367.41 seconds

[RESULT] BW   phase 3             ior_easy_read                8.090 GiB/s : time 273.37 seconds

[RESULT] BW   phase 4             ior_hard_read                3.726 GiB/s : time 111.69 seconds

[RESULT] IOPS phase 1         mdtest_easy_write                4.263 kiops : time 3518.34 seconds

[RESULT] IOPS phase 2         mdtest_hard_write                2.953 kiops : time 303.52 seconds

[RESULT] IOPS phase 3                      find               91.550 kiops : time 173.64 seconds

[RESULT] IOPS phase 4          mdtest_easy_stat              137.243 kiops : time 109.30 seconds

[RESULT] IOPS phase 5          mdtest_hard_stat               84.140 kiops : time  10.65 seconds

[RESULT] IOPS phase 6        mdtest_easy_delete               55.311 kiops : time 271.19 seconds

[RESULT] IOPS phase 7          mdtest_hard_read               21.778 kiops : time  41.16 seconds

[RESULT] IOPS phase 8        mdtest_hard_delete                7.133 kiops : time 129.24 seconds

[SCORE] Bandwidth 3.81212 GiB/s : IOPS 24.1149 kiops : TOTAL 9.58796

The io500.sh was run

 

Running the C version of the benchmark now

IO500 version io500-isc20

<currently running …when I posted this question ….

 

 

***************************************************

config-test1.ini (END)

***************************************************

[global]

datadir = ./out/

resultdir = ./results/

 

timestamp-resultdir = TRUE

 

# Chose parameters that are very small for all benchmarks

 

[debug]

stonewall-time = 300 # for testing

 

[ior-easy]

transferSize = 2m

blockSize = 102400m

 

[mdtest-easy]

API = POSIX

# Files per proc

n = 500000

 

[ior-hard]

API = POSIX

# Number of segments  10000000

segmentCount = 400000

 

[mdtest-hard]

API = POSIX

# Files per proc 1000000

n = 40000

 

[find]

external-script = /mnt/beeond/io500-app/bin/pfind

pfind-parallelize-single-dir-access-using-hashing = FALSE

 

***************************************************

 

_______________________________________________
IO-500 mailing list
IO-500@vi4io.org
https://www.vi4io.org/mailman/listinfo/io-500