vg0

Members
  • Posts

    4
  • Joined

  • Last visited

Posts posted by vg0

  1. Testresults for rc12a / rc12a /w SF 1.0.11.

    (not 100% baremetal, unmenu was installed on both tests)

     

    Test number 1.

     

    First i tested parity speeds on a stock rc12a, no SimpleFeatures, no other SF plugins, no 3rd party plugins, only unmenu.

    I've started a parity check and checked the average speed every 5 minutes, for about 30 minutes.

    In this case, the values shown by the stock gui are pretty much the same as within unmenu.

     

    minute: speed MB/s

    0: 166 MB/s

    5: 155 MB/s

    10: 155 MB/s

    15: 152 MB/s

    20: 145 MB/s

    25: 143 MB/s

    30: 142 MB/s

     

     

    Test number 2.

     

    This is unraid rc12a with only the SimpleFeatures 1.0.11 core component installed. No other SF plugins, no 3rd party plugins. System was rebooted.

    I've started the parcheck and waited for unmenu to give a stable reading (right column) after that i switched

    to SimpleFeatures and refreshed the page a few times so it would give a stable reading (left column).

    Notice that refreshing the SF gui makes the parity speed drop about 50%... switching to unmenu shows the same low speed.

    Closing the SF tab and watching unmenu, the speed is back to 100% in about 10-20 seconds.

     

    minute SF MB/s UM MB/s

    0: 98 MB/s 162 MB/s

    5: 85 MB/s 155 MB/s

    10: 85 MB/s 152 MB/s

    15: 82 MB/s 150 MB/s

    20: 72 MB/s 149 MB/s

    25: 73 MB/s 148 MB/s

    30: 71 MB/s 145 MB/s

     

    Also, if you take a look at the syslog, there are some disturbing CPU stall errors... dont know if that is related.

     

    Needles to say i am not going to test with others plugins, the core component itself is somehow disturbing the parity check speed, is my conclusion at the moment.

     

    Hi Jowi,

     

    Thanks for doing that testing. I'm of two minds, but will need to do some serious investigation. No way should a PHP function cause cpu stalls.

     

    Either  way, thankfully you've proved that the speed is the same unless you view the web gui. Unless you are refreshing the GUI for 8 hours straight you won't see a huge difference.

     

    I'll need to review the code to inspect what changes might be causing this slowdown. Whatever it will be, IT IS NOT DANGEROUS. It will simply be something that causes I/O interruption, something that browsing a share while parity check is running will do as well.

     

    Im having the a similar problem which i think is related to this.

    I noticed this when doing reads of large files from shares.

    Using the windows 8 file transfer graph, which is quite useful its easy to see the reads stop momentarily every 5 seconds for a short time before continuing,

    This is also visible on the network throughput monitor