sysbench-developers team mailing list archive

[dpfarrer <question254277@xxxxxxxxxxxxxxxxxxxxx>]

New question #254277 on sysbench:
https://answers.launchpad.net/sysbench/+question/254277

I have two identically outfitted Dell 2950 iii servers running Ubuntu 14.04: gb2a and gb2b.  Both systems have 32GB RAM and 6 2TB disks in a hardware RAID 6. Both have dual Xeon  X5450  @ 3.00GHz.

 When I time the i/o prepare command these are the results:

root@gb2b:~# time sysbench --test=fileio --file-total-size=150G prepare
sysbench 0.4.12:  multi-threaded system evaluation benchmark

128 files, 1228800Kb each, 153600Mb total
Creating files for the test...

real    11m44.316s
user    0m0.700s
sys     3m12.292s
root@gb2b:~#

root@gb2a:~# time sysbench --test=fileio --file-total-size=150G prepare
sysbench 0.4.12:  multi-threaded system evaluation benchmark

128 files, 1228800Kb each, 153600Mb total
Creating files for the test...

real    42m3.393s
user    0m8.691s
sys     39m52.104s
root@gb2a:~# ~

AS you can see system gb2a takes ~4 times longer to complete. When I run the actual test the results are nearly identical: 

root@gb2a:~# sysbench --test=fileio --file-total-size=150G --file-test-mode=rndrw --init-rng=on --max-time=300 --max-requests=0 run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1
Initializing random number generator from timer.


Extra file open flags: 0
128 files, 1.1719Gb each
150Gb total file size
Block size 16Kb
Number of random requests for random IO: 0
Read/Write ratio for combined random IO test: 1.50
Periodic FSYNC enabled, calling fsync() each 100 requests.
Calling fsync() at the end of test, Enabled.
Using synchronous I/O mode
Doing random r/w test
Threads started!
Time limit exceeded, exiting...
Done.

Operations performed:  31500 Read, 21000 Write, 67193 Other = 119693 Total
Read 492.19Mb  Written 328.12Mb  Total transferred 820.31Mb  (2.729Mb/sec)
  174.66 Requests/sec executed

Test execution summary:
    total time:                          300.5868s
    total number of events:              52500
    total time taken by event execution: 271.8952
    per-request statistics:
         min:                                  0.01ms
         avg:                                  5.18ms
         max:                               1326.14ms
         approx.  95 percentile:              11.21ms

Threads fairness:
    events (avg/stddev):           52500.0000/0.00
    execution time (avg/stddev):   271.8952/0.00

----------- and here is gb2b -----------------
root@gb2b:~# sysbench --test=fileio --file-total-size=150G --file-test-mode=rndrw --init-rng=on --max-time=300 --max-requests=0 run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1
Initializing random number generator from timer.


Extra file open flags: 0
128 files, 1.1719Gb each
150Gb total file size
Block size 16Kb
Number of random requests for random IO: 0
Read/Write ratio for combined random IO test: 1.50
Periodic FSYNC enabled, calling fsync() each 100 requests.
Calling fsync() at the end of test, Enabled.
Using synchronous I/O mode
Doing random r/w test
Threads started!
Time limit exceeded, exiting...
Done.

Operations performed:  33109 Read, 22072 Write, 70528 Other = 125709 Total
Read 517.33Mb  Written 344.88Mb  Total transferred 862.2Mb  (2.874Mb/sec)
  183.93 Requests/sec executed

Test execution summary:
    total time:                          300.0030s
    total number of events:              55181
    total time taken by event execution: 278.4508
    per-request statistics:
         min:                                  0.01ms
         avg:                                  5.05ms
         max:                               1252.96ms
         approx.  95 percentile:              10.79ms

Threads fairness:
    events (avg/stddev):           55181.0000/0.00
    execution time (avg/stddev):   278.4508/0.00

I ran sysbench in the first place to try and figure out why gb2a is so much slower than its identical twin. Does the slowness in file prep on gb2a tell you anything?

Many thanks,

DP


-- 
You received this question notification because you are a member of
sysbench-developers, which is an answer contact for sysbench.