Efficient low cost home made NAS

  • Hi,


    this is my first NAS, target was to build a custom device comparable to Synology ds216j in terms of cost and energy efficiency, but with better performance and extensibility.


    I choose ASRock J3455-ITX for cost, low-power and 4 SATA3 ports. Here the list of components:



    ItemPrice (EUR)
    Dealer
    be quiet! PC- Netzteil Be Quiet System Power B8 300W bulk Cod. BN25631,00Italian Internet Store
    MB Intel ASRock J3455-ITX M-ITX, 2xD3 1866 USB3 SATA3 Cod. 90-MXB3W0-A0UAYZ72,60Italian Internet Store
    memory SO D3 1600 4GB C11 Kingston 1x4GB Value Ram, 1,35/1,5V Cod. KVR16LS11/433,00Italian Internet Store
    Western Digital HDD int. 3,5 4TB WD WD40EZRZ, Blue SATA 6Gb/s 5400rpm 64MB Cod. WD40EZRZ92,80Italian Internet Store
    Italian Internet Store Shipping6,90Italian Internet Store
    Cooler Master Elite 130 case42,99Amazon
    Cooler Master Elite 130 case-5,00Amazon
    Sandisk Cruzer Edge 16GB (USB 2.0)8,00Local store
    TOTAL282,29




    I installed Debian 9.5 and latest OMV 4.1.10 (Arrakis) on USB stick.The only app is Tiny Tiny RSS customized by me.
    In the next future I would like to implement a backup system for devices in my family (laptops, Android mobiles, etc.).


    Consumption is very low:
    - Off: 0.5W (only PSU is on and WOL enabled)
    - On: 15-17W
    - Heavy load: not tested yet


    Every day I startup (WOL) and shutdown.


    MB BIOS configuration

    • Advanced

      • Chipset Configuration

        • Primary Graphics Adapter: PCI Express
        • Deep S5: Auto
      • ACPI Configuration

        • PCIE Devices Power On: Enabled
    • H/W Monitor

      • CPU Fan 1 Setting: Automatic mode

        • Target CPU Temperature: 50 °C/122 °F
        • Target Fan Speed: Level 1
      • Chassis Fan 1 Setting

        • Chassis Fan 1 Temp Source: Monitor M/B
        • Target CPU Temperature: 50 °C/122 °F
        • Target Fan Speed: Level 1



    Cheers from Italy

  • Heavy load: not tested yet

    Maybe we can make a 'deal'? :)


    I'm interested in raw CPU performance of those more energy efficient Intel boxes to get comparison numbers for https://github.com/ThomasKaiser/sbc-bench (currently I only having numbers for Atom x5-Z8300 and x5-Z8350 which are pretty low-end).


    When executing sbc-bench -c at least within the last 5 minutes when the cpuminer test is executed you should get a consumption number for your setup under full CPU under full load (of course without HDD activity but this is more ore less a matter of 'doing the math')


    Edit: on a 'plain' Debian most probably little Linux knowledge is needed since make and gcc might be missing. Also thermal sources (find /sys -name temp will reveal a lot of sensors) need adjustment.

  • Hi Thomas,


    I registered here to participate to this forum... so, deal accepted :)


    I run the script, during the execution it shows several errors at line 164:
    sbc-bench.sh: line 164: /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq: No such file or directory


    The output of the script has been uploaded to ix.io server, do you want me to share the URL with you?
    I didn't monitor the energy consumption, am I allowed to run the script whenever I want? I can exclude the upload of the data.


    Thanks and regards,
    Gabriele

  • Hope I'm allowed to answer as well ;)
    Have a similar system with a
    ASrock J4205 ITX
    1x8GB So-Dimm 8GB 1866
    64GB SSD
    PicoPSU 90
    60W Seasonic NT
    3 WD Red (1x8GB, 2x4GB)


    Power consumption whilst the miner test 16 Watts without fans, HDs incl. (of course ssd)



    Memory performance:
    memcpy: 5070.5 MB/s (1.1%)
    memset: 5170.6 MB/s



    Cpuminer total scores (5 minutes execution): 18.84,18.83,18.82,18.81,18.80,18.79,18.78,18.77,18.74,18.70,18.69,18.68,18.67,18.66,18.65,18.64,18.56 kH/s



    7-zip total scores (3 consecutive runs): 7578,7566,7302



    OpenSSL results:
    type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
    aes-128-cbc 355657.87k 553138.90k 614528.26k 656344.75k 667129.17k 669788.84k
    aes-128-cbc 355422.18k 553424.62k 616944.98k 657374.89k 669130.75k 670072.83k
    aes-192-cbc 321869.41k 437222.74k 520436.91k 550778.54k 558429.53k 558972.93k
    aes-192-cbc 322595.42k 440653.74k 521240.75k 551463.94k 559423.49k 559398.91k
    aes-256-cbc 295735.88k 390097.49k 451021.48k 473652.91k 480501.76k 480449.88k
    aes-256-cbc 296563.55k 393264.98k 452747.43k 470676.14k 479046.31k 480832.17k



    http://ix.io/1m5t

  • Ciao Gabriele,


    I run the script, during the execution it shows several errors at line 164:
    sbc-bench.sh: line 164: /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq: No such file or directory


    Interesting. Pretty 'common' problem I encountered now a few times when tests ran on x86.'cpuinfo_cur_freq' contains the 'real' clockspeed while 'scaling_cur_freq' contains 'the frequency the kernel thinks the CPU runs at'. You could change line 164 of the script and replace cpuinfo_cur_freq with the other most probably existing sysfs node but it won't help that much anyway since on TurboBoost enabled CPUs reported and real clockspeeds still don't need to match.


    But that's what the other measurements ensure, see @Stramm numbers (thanks!).


    Yeah, please share the URL and of course you can run the tool at any time. But if it's just about to generate a lot of heat and maybe check if cooling is appropriate the '-T' and '-t' modes are better. They always use the most demanding benchmark and are designed to test for 'thermal efficiency', see this example.

  • Hi,


    here the output file
    http://ix.io/1m5p


    and below totals.


    @Stramm, your configuration is very interesting, 4205 was a serious option, but was not in my budget. I will comment later.


    Memory performance:
    standard memcpy : 4086.8 MB/s
    standard memset : 4045.5 MB/s (0.1%)



    Cpuminer total scores (5 minutes execution):
    17.30,17.28,17.27,17.26,17.25,17.24,17.22,17.21,17.20,17.18,17.17,17.16,17.15,17.14,17.13,17.12,17.11,17.10,17.09,17.08,17.07,17.06,17.05,17.03,17.02,17.01,17.00,16.99,16.95,16.90,16.58,16.56



    7-zip total scores (3 consecutive runs):
    7003,7003,7008



    OpenSSL results:
    type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
    aes-128-cbc 316649.82k 494317.61k 549095.85k 586563.93k 597407.06k 597934.08k
    aes-128-cbc 316304.14k 490763.37k 549233.07k 586658.47k 596929.19k 598491.14k
    aes-192-cbc 281071.98k 387220.80k 464883.46k 491654.83k 499324.25k 500312.75k
    aes-192-cbc 288168.99k 386304.62k 461889.19k 492381.53k 499067.56k 499733.85k
    aes-256-cbc 242770.45k 335505.07k 402155.35k 423486.81k 429331.80k 429735.94k
    aes-256-cbc 263030.98k 349985.56k 403909.21k 423625.39k 429400.06k 429583.02k



  • I have to admit that I bought the 4205 cause the 3455 was out of stock. Usually patience is not one of my strength so I swallowed the pill and invested 15€ more. And the little advantage of the 4205 isn't worth the money.


    Edit: I see the difference in Italy is even bigger. So it's totally pointless to go for the 4205.

Jetzt mitmachen!

Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!