Home Forums Software discussion Performance Tests

Tagged: 

This topic contains 2 replies, has 2 voices, and was last updated by  bunny 4 days, 18 hours ago.

Viewing 3 posts - 1 through 3 (of 3 total)
  • Author
    Posts
  • #1264

    bunny
    Participant

    Im using my espressobin as a NAS server. I was curious about the I/O and network performance.
    Here my findings.

    OS: Ubuntu 16.04.3 LTS
    Network Share: Samba
    Connected Storage: USB 2.0 ext4 formatted 128GB USB Flash drive
    USB 3.0 ntfs formatted 2TB HDD

    Network connection: Test PC and Espressobin connected to 1GBit Ports on my Router

    Network speed test with iPerf3

    Iperf		write
    [ ID] Interval           Transfer     Bandwidth       Retr
    [  4]   0.00-60.00  sec  5.50 GBytes   787 Mbits/sec   24             sender
    [  4]   0.00-60.00  sec  5.49 GBytes   786 Mbits/sec                  receiver

    –> ~98 MB/s

    Write Tests:
    1. dd if=/dev/zero of=/mnt/espressoUSB/tst bs=10M count=10

    Results:
    USB 2.0 ext4 formatted 128GB USB Flash drive:
    524288000 bytes (524 MB, 500 MiB) copied, 22,9038 s, 22,9 MB/s
    –> 22,9 MB/s

    USB 3.0 ntfs formatted 2TB HDD
    524288000 bytes (524 MB, 500 MiB) copied, 28,5761 s, 18,3 MB/s
    –> 18,3 MB/s

    2. Copy of a test 1GB file
    rsync --stats --progress 1GB.zip /mnt/espressoUSB/

    Results:
    USB 2.0 ext4 formatted 128GB USB Flash drive:
    sent 1,074,004,047 bytes received 35 bytes 21,697,052.16 bytes/sec
    –>20,7 MB/s

    USB 3.0 ntfs formatted 2TB HDD
    sent 1,074,004,047 bytes received 35 bytes 18,050,488.77 bytes/sec
    –>17,2 MB/s

    Which gives me the impression that the network speed is really great, but USB 3.0 not really needed, since the write speed is ~20 MB/s so USB 2.0 is more then enough.

    Does my tests matches also your results? I’m curious to see if you managed to tweak your espressobin so you can get better results.

    /Bunny

    #1296

    tkaiser
    Participant

    Testing with and using NTFS is no good idea, the next time use htop to see the CPU core being busy handling this foreign filesystem through FUSE bottlenecked by the slow CPU and running single-threaded.

    If you repeat the test with a better suited filesystem on USB3 you should see close to native HDD speeds (even the older Armada 38x is able to exceed 250 MB/s via USB3 without any problem). Though I’ve no idea why one wants to do performance tests on USB ports on a system with that performant native SATA (+500 MB/s and also high random IO, though that’s something for SSDs then)

    #1300

    bunny
    Participant

    I changed my setup according to recommendations from tkaiser. I see a big performance change when using dd, but with real data (see copy of a 1GB test file) i somehow hit the limit of 30 MB/s.

    1. dd if=/dev/zero of=/mnt/espressoUSB/tst bs=50M count=10

    Results:
    USB 3.0 ext4 formatted 128GB USB Flash drive:
    524288000 bytes (524 MB, 500 MiB) copied, 8,90604 s, 58,9 MB/s

    USB 2.0 ntfs formatted 2TB HDD
    524288000 bytes (524 MB, 500 MiB) copied, 18,0389 s, 19,1 MB/s

    SATA ext4 formatted 2TB HDD
    524288000 bytes (524 MB, 500 MiB) copied, 4,84963 s, 108 MB/s

    2. Copy of a test 1GB file

    Results:
    USB 3.0 ext4 formatted 128GB USB Flash drive:
    sent 1,074,004,046 bytes received 35 bytes 20,854,448.17 bytes/sec
    –> 25 MB/s

    USB 2.0 ntfs formatted 2TB HDD
    sent 1,074,004,046 bytes received 35 bytes 8,912,896.94 bytes/sec
    –> 8,5 MB/s

    SATA ext4 formatted 2TB HDD
    sent 1,074,004,046 bytes received 35 bytes 23,604,485.30 bytes/sec
    –> 22,5 MB/s

    SATA ntfs formatted 2TB HDD
    sent 1,074,004,046 bytes received 35 bytes 14,225,219.62 bytes/sec
    –> 13,6 MB/s

Viewing 3 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic.

Signup to our newsletter

Technical specification tables can not be displayed on mobile. Please view on desktop