Slow SMB transfer speeds

1. offizieller Beitrag

    So i've just bought an HP Microserver Gen8. Installed an SSD in the ODD bay, and configured it to be bootable. I run OMV from this drive.


    When I transfer files to and from the server, I barely go up to 9 MB/s.


    I've got a TP-Link AX50 router, gen8 connected via cat 6a cables (not necessary I know) and windows client connected via 802.11n.


    So I should be able to reach closer to 50-60 MB/s on 802.11n?


    Pinged router and server from windows client, and vice verse with server. Get 0% loss.


    What could be the issue?

    Maybe by now you had a bit time to read pinned threads and found the SMB performance thread already,

    In case not, the link is in my signature below.

    omv 6.9.6-2 (Shaitan) on RPi CM4/4GB with 64bit Kernel 6.1.21-v8+

    2x 6TB 3.5'' HDDs (CMR) formatted with ext4 via 2port PCIe SATA card with ASM1061R chipset providing hardware supported RAID1


    omv 6.9.3-1 (Shaitan) on RPi4/4GB with 32bit Kernel 5.10.63 and WittyPi 3 V2 RTC HAT

    2x 3TB 3.5'' HDDs (CMR) formatted with ext4 in Icy Box IB-RD3662-C31 / hardware supported RAID1

    For Read/Write performance of SMB shares hosted on this hardware see forum here

    • Offizieller Beitrag
    Zitat von mi-hol

    Maybe by now you had a bit time to read pinned threads and found the SMB performance thread already,

    Man! I don’t even know what that means, or even how it could help. If you don’t have real advise don’t respond to the post.

    System Backup Typo alert: Under the Linux section the command should be sudo umount /dev/sda1 NOT sudo unmount /dev/sda1

    Backup Data Disk to Backup Disk on Same Machine: In a Scheduled Job:rsync -av --delete /srv/dev-disk-by-uuid-f8814ed9-9a5c-4e1c-8830-426968c20ea3/ /srv/dev-disk-by-uuid-e67439d5-00a3-4942-bd5f-b84ab86aa850/ Don't forget trailing slashes, and BE CAREFUL. (HT: Getting Started with OMV5)

    Equipment - Thinkserver TS140, NanoPi M4 (v.1), Odroid XU4 (Using DietPi): PiHole

    Gefällt mir 1
    • Offizieller Beitrag
    Zitat von Neomorph

    So I should be able to reach closer to 50-60 MB/s on 802.11n?

    You should do better than 9 but there are a number of factors involved.


    First, we're talking about wireless. You'd need to check the wireless speed at which you're connecting to the router. (Is it a Laptop?) Realize, in most cases, a radio link is a simplex connection which means when overhead is exchanged that cuts data transfer speed somewhat.

    Second, if you live in an apartment or close to a neighboring house, there may be interference on the router's channel.


    The first thing I would do is test with the windows client hard wired to the router.
    See what that looks like.

    Zitat von Agricola

    I don’t even know what that means

    in simple terms it means, the 9Mb/s is perfectly ok, behause it falls within the measured speeds for 802.11n in the post (copied below).


    I'm surprised its not so simple to understand but look forward to suggests how this can be improved.


    2) Connected via WiFi/WLAN(802.11n) (WiFi router was old single core model)

    [Read]

    SEQ 1MiB (Q= 8, T= 1): 6.1 MB/s [ 5.8 IOPS]

    SEQ 1MiB (Q= 1, T= 1): 5.9 MB/s [ 5.6 IOPS]

    RND 4KiB (Q= 32, T= 1): 5.7 MB/s [ 1395.3 IOPS]

    RND 4KiB (Q= 1, T= 1): 1.3 MB/s [ 313.5 IOPS]


    [Write]

    SEQ 1MiB (Q= 8, T= 1): 11.7 MB/s [ 11.2 IOPS]

    SEQ 1MiB (Q= 1, T= 1): 10.1 MB/s [ 9.6 IOPS]

    RND 4KiB (Q= 32, T= 1): 9.7 MB/s [ 2356.0 IOPS]

    RND 4KiB (Q= 1, T= 1): 1.3 MB/s [ 317.6 IOPS]


    3) Connected via WiFi/WLAN (802.11n) (WiFi router was new dual core model)

    [Read]

    SEQ 1MiB (Q= 8, T= 1): 14.2 MB/s [ 13.6 IOPS]

    SEQ 1MiB (Q= 1, T= 1): 13.6 MB/s [ 13.0 IOPS]

    RND 4KiB (Q= 32, T= 1): 11.3 MB/s [ 2765.9 IOPS]

    RND 4KiB (Q= 1, T= 1): 2.0 MB/s [ 485.6 IOPS]


    [Write]

    SEQ 1MiB (Q= 8, T= 1): 12.2 MB/s [ 11.6 IOPS]

    SEQ 1MiB (Q= 1, T= 1): 11.3 MB/s [ 10.8 IOPS]

    RND 4KiB (Q= 32, T= 1): 10.9 MB/s [ 2655.3 IOPS]

    RND 4KiB (Q= 1, T= 1): 1.9 MB/s [ 458.0 IOPS]

    omv 6.9.6-2 (Shaitan) on RPi CM4/4GB with 64bit Kernel 6.1.21-v8+

    2x 6TB 3.5'' HDDs (CMR) formatted with ext4 via 2port PCIe SATA card with ASM1061R chipset providing hardware supported RAID1


    omv 6.9.3-1 (Shaitan) on RPi4/4GB with 32bit Kernel 5.10.63 and WittyPi 3 V2 RTC HAT

    2x 3TB 3.5'' HDDs (CMR) formatted with ext4 in Icy Box IB-RD3662-C31 / hardware supported RAID1

    For Read/Write performance of SMB shares hosted on this hardware see forum here

    Well the OP has conflicting information.

    802.11ac and 802.11n are mentioned.

    So lets wait for answer from OP

    omv 6.9.6-2 (Shaitan) on RPi CM4/4GB with 64bit Kernel 6.1.21-v8+

    2x 6TB 3.5'' HDDs (CMR) formatted with ext4 via 2port PCIe SATA card with ASM1061R chipset providing hardware supported RAID1


    omv 6.9.3-1 (Shaitan) on RPi4/4GB with 32bit Kernel 5.10.63 and WittyPi 3 V2 RTC HAT

    2x 3TB 3.5'' HDDs (CMR) formatted with ext4 in Icy Box IB-RD3662-C31 / hardware supported RAID1

    For Read/Write performance of SMB shares hosted on this hardware see forum here

    • Offizieller Beitrag
    Zitat von mi-hol

    802.11ac and 802.11n are mentioned.

    Yes best guess would be, is either the router is ac and the laptop n or vice versa, but as crashtest test mentioned 2.4GHz is highly susceptible to interference unlike 5GHz


    EDIT: you can also add to that, distance from source to router, thereby placing walls etc 'in the way'

    Agricola i've added below to the post, does this help?

    Short summary:

    • A RPi4 will max out a 1Gbit/s network (common maximum speed in home and office networks)
    • performance bottleneck for clients accessing data on the NAS is the used network
      • best case for wireless network access via WiFi-4/801.11n is for Write ~ 10 MByte/s & Read ~ 13.5 MByte/s
      • best case for wireless network access via WiFi-5/801.11ac is Write ~ 18 MByte/s & Read ~ 20 MByte/s
      • random access of data is much slower between 1,3 MByte/s to 2 MByte/s for WiFi-4/801.11n & WiFi-5/801.11ac
    • investing in fast NAS hardware has low to no benefit to clients access the data, behause network ist the real bottleneck

    omv 6.9.6-2 (Shaitan) on RPi CM4/4GB with 64bit Kernel 6.1.21-v8+

    2x 6TB 3.5'' HDDs (CMR) formatted with ext4 via 2port PCIe SATA card with ASM1061R chipset providing hardware supported RAID1


    omv 6.9.3-1 (Shaitan) on RPi4/4GB with 32bit Kernel 5.10.63 and WittyPi 3 V2 RTC HAT

    2x 3TB 3.5'' HDDs (CMR) formatted with ext4 in Icy Box IB-RD3662-C31 / hardware supported RAID1

    For Read/Write performance of SMB shares hosted on this hardware see forum here

    Zitat von geaves

    best guess would be, is either the router is ac and the laptop n

    correct and therefore an expectation of "able to reach closer to 50-60 MB/s on 802.11n" is plain simply wrong.


    So my approach is instead of having basically the same conversation again and again, help to set realistic expectations and document guidance how everyone is able to achieve a certain level of performance by following the guidance.

    omv 6.9.6-2 (Shaitan) on RPi CM4/4GB with 64bit Kernel 6.1.21-v8+

    2x 6TB 3.5'' HDDs (CMR) formatted with ext4 via 2port PCIe SATA card with ASM1061R chipset providing hardware supported RAID1


    omv 6.9.3-1 (Shaitan) on RPi4/4GB with 32bit Kernel 5.10.63 and WittyPi 3 V2 RTC HAT

    2x 3TB 3.5'' HDDs (CMR) formatted with ext4 in Icy Box IB-RD3662-C31 / hardware supported RAID1

    For Read/Write performance of SMB shares hosted on this hardware see forum here

    I wont use that word :)


    Most people are attracted by doing /finding something new, repeating things bores and exhausts most people and is hence better left to computers

    omv 6.9.6-2 (Shaitan) on RPi CM4/4GB with 64bit Kernel 6.1.21-v8+

    2x 6TB 3.5'' HDDs (CMR) formatted with ext4 via 2port PCIe SATA card with ASM1061R chipset providing hardware supported RAID1


    omv 6.9.3-1 (Shaitan) on RPi4/4GB with 32bit Kernel 5.10.63 and WittyPi 3 V2 RTC HAT

    2x 3TB 3.5'' HDDs (CMR) formatted with ext4 in Icy Box IB-RD3662-C31 / hardware supported RAID1

    For Read/Write performance of SMB shares hosted on this hardware see forum here

    So, to update a bit of information!

    The windows client I'm using is a stationary PC, with a D-Link DWA-556 card (802.11n), and is on the floor above the router. (this would explain the poor transfer speeds, but I'm getting close to 11MB/s with internet downloads from the same client.)


    I live in a house with no surrounding interference from neighbors


    The router, as mentioned, is an 802.11ax router. I have since the last post tried with a laptop, with an 802.11ax card installed, right next to the router, and maxing out at around 30-35 MB/s, which is quite poor for that standard, no?


    I will conduct a proper "benchmark" during the week when my newly ordered laptop (again with 802.11ax card) will arrive. And that will include:


    - wireless transfer from the top floor.

    - wireless transfer next to the router.

    - wired transfer with gigabit ethernet connected directly to the router.


    All file transfers are, again, done through SMB. The goal is for the NAS to be used as a Plex server, and I have not yet configured anything doing with Plex, as I want to try and understand/possible fix the poor transfer speeds.

    • Offizieller Beitrag
    Zitat von Neomorph

    The router, as mentioned, is an 802.11ax router. I have since the last post tried with a laptop, with an 802.11ax card installed, right next to the router, and maxing out at around 30-35 MB/s, which is quite poor for that standard, no?

    When using wireless and testing for throughput, there's a number of factors involved that won't have anything to do with Samba (SMB). If you're running wireless and trying to determine the best locations in your house for the best speed, etc., you might consider taking a look at -> iperf. Iperf works in a client / server arrangement. Install Iperf on two different clients for point-to-point testing with two wired connections to the router, one wired with the other wireless, etc.


    Iperf will give you an idea of what your best case transfer rate is, wired or wireless. You'll also get an indication of how well your wireless links are doing, the locations in your house that are best, local interference factors, etc.

    You've said that you don't live close to others but you might want to see if your router has features that will either auto select the best channel(s) with the least interference or a method of letting you see the radio spectrum to make manual channel choices. (Note that interference can "local", in the 2.4Ghz band. 5Ghz is less susceptible, due to fewer consumer devices using it.)


    Once you have this information in hand, then looking at SMB transfer rates might be a bit more meaningful.


    I will say, as you're about to discover, theoretical maximums (802.11ax) are rarely seen in the real world.

    I've run iperf between my windows client and the server, as well as between the windows client and a secondary laptop (then both on wireless) and have received 0% loss with both. To add is I've not changed any parameters, and just run iperf as it is.

    I will test it again once, as mentioned, my laptop will arrive. Then I'll test both wired and wireless from different point in the house, to locate blind spots within the network.


    Is there anything more to be read from iperf than how much data loss there is between client and server?


    And yes, my router has a "Smart Connect" feature, where it combines 2,4GHz and 5GHz with one SSID. and chooses the best for band for the fastest transfer speeds. My windows client seems to be on the 2,4GHz band though.

    • Offizieller Beitrag
    Zitat von Neomorph

    And yes, my router has a "Smart Connect" feature, where it combines 2,4GHz and 5GHz with one SSID. and chooses the best for band for the fastest transfer speeds

    My router has the same option but I choose to separate each channel, depending on the hardware and where it's likely to be used will determine which band the hardware is set to. As to Smart Connect, I turn mine off.

    • Offizieller Beitrag

    Look at iperf's doc's. While no loss of packets is good, you're not looking for loss. (Data loss would be bad news.)

    You're looking at "bandwidth" in "k" (kilo) or "m" (meg). You can set the measurement standard. What you're interested in is "throughput" as in, what you're getting over the air versus through a wired connection. "Through the wire" will be the reference (gold) standard, used to compare to wireless performance.


    Here's a short iperf -> tutorial.


    I'd test:

    wired to wired (through the router)

    wired to wireless (to various locations in house looking for bad or dead spots.)

Jetzt mitmachen!

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

Benutzerkonto erstellen Anmelden