Looking to take advantage of 10GbE and overcome HD bottleneck

  • Just upgraded my home lab to 10GbE, picked up a cheap 10GbE NIC... and I'm looking to reduce the hard drive speeds as a bottleneck.


    Currently I don't have raid, and my SAMBA shares act as a landing zone for Greyhole to run in the background and distribute files on other drives for redundancy.


    Since this home NAS is a bit of DIY fun, I'm looking at cheap and creative ways to take advantage of 10GbE,

    1. Could I add a RAID / SSD array as a cache to frequently accessed files?
    2. Or is my best bet to put the SAMBA shares on a more classic RAID, which doubles as the Greyhole landing zone. I'm not sure how this will affect read speeds, but I guess if I make the RAID LZ big enough, and also a destination drive, most reads will come off that drive?
    3. Or is it time for a rebuild? I love Greyhole and its ability to pool storage across random sized drives. But if it's not going to be performant I may look to change things.
    4. Finally, are there any resources on how I can optimise Linux with my build in mind for multi GB LANs?

    CPU i7-2600 | 11 GB RAM | Mobo Asus P8Z68-V LX | HP NC550SFP 10GbE 2-Port PCI-E-2.0x8 (SFP+ DAC) 802.3ad Bonding | US-16-XG + UDM Pro

    • OMV 6.4.6-1 (Shaitan). Installed and booting via USB 3 SATA SSD
    • SAMBA shares, doubling as landing zone for Greyhole redundant storage pool file system service.
    • 3.25 Drives (in TB) - 1,1,1,0.3,2 - 2TB drive dedicated to Timemachine.
  • square_eyes

    Hat den Titel des Themas von „Looking to take advantage of 10GbE“ zu „Looking to take advantage of 10GbE and overcome HD bottleneck“ geändert.
  • To answer much more info about your hardware is needed.

    Best in the signature

    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

  • Updated, and thanks for the reply :)

    CPU i7-2600 | 11 GB RAM | Mobo Asus P8Z68-V LX | HP NC550SFP 10GbE 2-Port PCI-E-2.0x8 (SFP+ DAC) 802.3ad Bonding | US-16-XG + UDM Pro

    • OMV 6.4.6-1 (Shaitan). Installed and booting via USB 3 SATA SSD
    • SAMBA shares, doubling as landing zone for Greyhole redundant storage pool file system service.
    • 3.25 Drives (in TB) - 1,1,1,0.3,2 - 2TB drive dedicated to Timemachine.
  • So what performance are you currently getting?

    Details on how to measure performance are in the post linked in my signature

    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

    Without raid, you won't get better than single disk performance. Since you are using all spinning disk, you are unlikely to break the 190MB/s mark. An SSD (cache or not) would help you get to about 500 MB/s. Since you only have three disks of the same size, you would probably struggle to hit 500 MB/s even in raid 0. raid 5 would struggle to hit 350 MB/s. All of those speeds are assuming your current disks can hit 180-190MB/s each.


    For a while, I ran a seven drive raid 0 array of 1TB disks on a hardware raid controller and it would do about 850MB/s at best.


    nvme is the easiest way to saturate 10gbe and doesn't require raid.

    omv 7.0.5-1 sandworm | 64 bit | 6.8 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.13 | compose 7.1.4 | k8s 7.1.0-3 | cputemp 7.0.1 | mergerfs 7.0.4


    omv-extras.org plugins source code and issue tracker - github - changelogs


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • the MOBO doesn't provide a NVMe interface, only:

    SATA-600 -connectors: 2 x 7pin Serial ATA - 2 devices - RAID 0 / RAID 1 / RAID 10 / RAID 5


    SATA-300 -connectors: 4 x 7pin Serial ATA - 4 devices - RAID 0 / RAID 1 / RAID 10 / RAID 5

    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

    the MOBO doesn't provide a NVMe interface, only:

    Uh, it has pci-e slots and a pci-e to nvme adapter is only about $10 on Amazon. I have them in a couple of my systems.

  • Thanks both! I actually have a single PCI to NVMe riser spare. That sounds like an option. Would a cache work - does OMV support caching?


    I have seen multi/raid NVMe PCIe risers but not sure if my old mobo supports those - something about lane splitting?

    CPU i7-2600 | 11 GB RAM | Mobo Asus P8Z68-V LX | HP NC550SFP 10GbE 2-Port PCI-E-2.0x8 (SFP+ DAC) 802.3ad Bonding | US-16-XG + UDM Pro

    • OMV 6.4.6-1 (Shaitan). Installed and booting via USB 3 SATA SSD
    • SAMBA shares, doubling as landing zone for Greyhole redundant storage pool file system service.
    • 3.25 Drives (in TB) - 1,1,1,0.3,2 - 2TB drive dedicated to Timemachine.
    • Offizieller Beitrag

    Would a cache work - does OMV support caching?

    Any caching would have to be setup from the command line.


    I have seen multi/raid NVMe PCIe risers but not sure if my old mobo supports those - something about lane splitting?

    Depends on if the pcie switch is on the card or not. Most don't have the switch and are expensive if they do. The cards without it needs specific motherboards usually. nvme doesn't need raid to saturate 10gbe though. I only have one card in my system (Samsung 960 Pro).

    omv 7.0.5-1 sandworm | 64 bit | 6.8 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.13 | compose 7.1.4 | k8s 7.1.0-3 | cputemp 7.0.1 | mergerfs 7.0.4


    omv-extras.org plugins source code and issue tracker - github - changelogs


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • nvme doesn't need raid to saturate 10gbe though

    Nice. I think this is the entry for me then. That just leaves...

    Any caching would have to be setup from the command line.

    Any services you can suggest that will play nice with SAMBA/Greyhole?

    CPU i7-2600 | 11 GB RAM | Mobo Asus P8Z68-V LX | HP NC550SFP 10GbE 2-Port PCI-E-2.0x8 (SFP+ DAC) 802.3ad Bonding | US-16-XG + UDM Pro

    • OMV 6.4.6-1 (Shaitan). Installed and booting via USB 3 SATA SSD
    • SAMBA shares, doubling as landing zone for Greyhole redundant storage pool file system service.
    • 3.25 Drives (in TB) - 1,1,1,0.3,2 - 2TB drive dedicated to Timemachine.
    • Offizieller Beitrag

    Any services you can suggest that will play nice with SAMBA/Greyhole?

    Caching is typically at the filesystem level and samba/greyhole would just see a normal filesystem. Using bcache or zfs with cache are pretty common. I use just the nvme drive with ext4 since it is a 1TB nvme drive. Keeps it simple and would work well with samba.

    omv 7.0.5-1 sandworm | 64 bit | 6.8 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.13 | compose 7.1.4 | k8s 7.1.0-3 | cputemp 7.0.1 | mergerfs 7.0.4


    omv-extras.org plugins source code and issue tracker - github - changelogs


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • I'm struggling to find a tutorial for optimising for 10GbE. I found this which has some steps but no real explanation as to where each setting is and what it's doing (such as - Parameters for NFS export: "async,no_subtree_check,insecure"). In another 10GbE thread there is a vague reference to OMV taking over all system config, and eventually overwriting things.


    Is there an approach to take with OMV in mind - e.g. omv-env variables and other system settings? I'd say with multi gig becoming more commonplace, it's worth a look if it doesn't exist somewhere already.

    CPU i7-2600 | 11 GB RAM | Mobo Asus P8Z68-V LX | HP NC550SFP 10GbE 2-Port PCI-E-2.0x8 (SFP+ DAC) 802.3ad Bonding | US-16-XG + UDM Pro

    • OMV 6.4.6-1 (Shaitan). Installed and booting via USB 3 SATA SSD
    • SAMBA shares, doubling as landing zone for Greyhole redundant storage pool file system service.
    • 3.25 Drives (in TB) - 1,1,1,0.3,2 - 2TB drive dedicated to Timemachine.

    Einmal editiert, zuletzt von square_eyes ()

    • Offizieller Beitrag

    I'm struggling to find a tutorial for optimising for 10GbE. I found this which has some steps but no real explanation as to where each setting is and what it's doing (such as - Parameters for NFS export: "async,no_subtree_check,insecure"). In another 10GbE thread there is a vague reference to OMV taking over all system config, and eventually overwriting things.


    Is there an approach to take with OMV in mind - e.g. omv-env variables and other system settings? I'd say with multi gig becoming more commonplace, it's worth a look if it doesn't exist somewhere already.

    What are you trying to optimize? nfs shouldn't need any optimization based on the network adapter speed. I use the default settings for nfs.

    omv 7.0.5-1 sandworm | 64 bit | 6.8 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.13 | compose 7.1.4 | k8s 7.1.0-3 | cputemp 7.0.1 | mergerfs 7.0.4


    omv-extras.org plugins source code and issue tracker - github - changelogs


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • vague reference to OMV taking over all system config

    that is incorrect, starting with OMV 5.6.13 all config files managed by OMV [update]core and core plugins are registered as "deviation" see https://github.com/openmediavault/openmediavault/issues/986

    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

    3 Mal editiert, zuletzt von mi-hol ()

    • Offizieller Beitrag

    starting with OMV 5,6,13 all config files managed by OMV are registered as deviations see

    Not quite true either. None of the omv-extras plugins register deviations. In general it is just easier to say OMV completely manages the services it controls. If you install transmission or apache, OMV will never touch either since there is no core OMV control or plugin for them.

    omv 7.0.5-1 sandworm | 64 bit | 6.8 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.13 | compose 7.1.4 | k8s 7.1.0-3 | cputemp 7.0.1 | mergerfs 7.0.4


    omv-extras.org plugins source code and issue tracker - github - changelogs


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • Apologize you are right, my statement was for OMV core and core plugins only

    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

Jetzt mitmachen!

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