What's new

How to improve Samba speed on Merlin 386.5 AX86U

  • SNBForums Code of Conduct

    SNBForums is a community for everyone, no matter what their level of experience.

    Please be tolerant and patient of others, especially newcomers. We are all here to share and learn!

    The rules are simple: Be patient, be nice, be helpful or be gone!

csmanul

Occasional Visitor
I recently bought a new ax86u flashed with merlin 386.5 , attached a wd blue ssd (WDS100T2B0B) enclosed in an USB 3.1 rack (orico M2L2-NV03C3-GY) then wired connected to my pc i7-8700 MX500 ssd with a CAT6 cable. i formatted the router drive ext4 no journalling then tried to do a file transfer to see if my investment works as advertised in numerous benchmarks and blogs praising the mighty ax86u having file transfer speeds over 100MB\s. I only get 55MB\s read or write no speed fluctuations tried in 3 operating systems (win,ubuntu,osx) speeds are identical, also as a precaution on unix based oses that i have i tested the transfer speed with rsync command ( with --progress option).
I understand that its not possible to match the benchmarks but still from 100MB\s+ to 50MB\s its a huge diff, something is wrong somewhere, my best bet is that the devs of merlin thought useful to limit the transfer speeds to avoid any issues with 2.4ghz interference.

TLDR: simple problem definition: on ftp i get 100MB/s on samba 50MB/s why ?
 
Last edited:
I forgot to mention that when transferring the router cpu is at about 25-28% in the web gui interface and roughly the same while using top unix cmd.
 
Your best bet is simply wrong. The performance of the hardware is identical to stock/RMerlin firmware (he doesn't touch closed source blobs).

What 'numerous' benchmarks suggest 100MB/s transfer speeds with real files?

What method are you specifically using to test these claims?

A single (large) file may hit those speeds (peak). A bunch of smaller files may not.
 
Your best bet is simply wrong. The performance of the hardware is identical to stock/RMerlin firmware (he doesn't touch closed source blobs).

What 'numerous' benchmarks suggest 100MB/s transfer speeds with real files?

What method are you specifically using to test these claims?

A single (large) file may hit those speeds (peak). A bunch of smaller files may not.
i'll give one review i knew on top of my head now, https://dongknows.com/asus-rt-ax86u-ax5700-gaming-router-review/ but i've seen other also with similar results, and on several forums including this one and product comments users report +100MB/s speeds.
i think i described how i tested the transfer speed well enough, in short all os'es copy/paste + unix based os'es rsync.
i am testing with a 4gb file, there is no peak or bottom the speed is fixed and stable.
 
running the following on router
Bash:
pv /dev/zero > /tmp/mnt/rack1tb/dummytestfile.txt
yields about 350-400MB\s
 
after further tinkering i narrowed down to the Samba protocol, a ftp transfer to the router usb hdd yields 114MB\s (decent, as expected) router cpu at 88%, so there is something wrong with samba.
 
Nope, nothing wrong with Samba. And Dongknows does not know everything. It is a simple fact that you should not use a router for a NAS. Period! Yes, one can use the USB port for file storage as long as you do not care about file transfer speed. Even a USB3 drive attached to my Synology has trouble getting to 100 MBPS.
The AX86U is still a great router...
 
Nope, nothing wrong with Samba. And Dongknows does not know everything. It is a simple fact that you should not use a router for a NAS. Period! Yes, one can use the USB port for file storage as long as you do not care about file transfer speed. Even a USB3 drive attached to my Synology has trouble getting to 100 MBPS.
The AX86U is still a great router...
i understand that you are a big fan but let's stick to the facts, do read my previous messages.
 
Have you looked at samba performance tuning options ...block sizes and options? Maybe the defaults are conservative? Stumped how the reviews could get double though as they would never delve into the samba configs..
 
Have you looked at samba performance tuning options ...block sizes and options? Maybe the defaults are conservative? Stumped how the reviews could get double though as they would never delve into the samba configs..
yep i have been trying some parameters like
Code:
socket options = TCP_NODELAY IPTOS_LOWDELAY SO_RCVBUF=131072 SO_SNDBUF=131072
send/recv buffers seem to have an impact if configured to small values => speed decreases, but at high values no changes still 50MB/s constant speed.
 
which versions of the SAMBA protocol are being used on each end ?
the client sees the server as smb2
on the router smbd 3.6.25 is running
Code:
% smbutil statshares -a

==================================================================================================
SHARE                         ATTRIBUTE TYPE                VALUE
==================================================================================================
rack1tb                     
                              SERVER_NAME                   192.168.0.1
                              USER_ID                       501
                              SMB_NEGOTIATE                 SMBV_NEG_SMB1_ENABLED
                              SMB_NEGOTIATE                 SMBV_NEG_SMB2_ENABLED
                              SMB_NEGOTIATE                 SMBV_NEG_SMB3_ENABLED
                              SMB_VERSION                   SMB_2.002
                              SMB_SHARE_TYPE                DISK
                              SIGNING_SUPPORTED             TRUE
                              EXTENDED_SECURITY_SUPPORTED   TRUE
                              LARGE_FILE_SUPPORTED          TRUE
                              FILE_IDS_SUPPORTED            TRUE

--------------------------------------------------------------------------------------------------
 
Last edited:
FYI: Using an old mechanical HDD attached to my RT-AX86U my Samba speeds are, Write: 112MB/s (i.e. maximum gigabit) and Read: 86MB/s.

BTW I couldn't find anything to suggest your Orico M2L2-NV03C3-GY is a "rack". It appears to be just a normal SDD to USB enclosure.
 
SMH another router/NAS thread.

At least you figured out the issue is SMB related by testing with FTP and getting the full speeds. SSD is overkill for USB / router / nas use though due to the cap on 1gig speeds.

If you truly want to hit speeds you need to get / DIY a NAS and hook it up to a switch @ higher than 1GE speeds. Plenty of 2.5/5/10GE options now that aren't insanely priced. My DIY network setup can hit close to 500MB/s over Ethernet and ~1.5gbps over WIFI.
 
SMH another router/NAS thread.

At least you figured out the issue is SMB related by testing with FTP and getting the full speeds. SSD is overkill for USB / router / nas use though due to the cap on 1gig speeds.

If you truly want to hit speeds you need to get / DIY a NAS and hook it up to a switch @ higher than 1GE speeds. Plenty of 2.5/5/10GE options now that aren't insanely priced. My DIY network setup can hit close to 500MB/s over Ethernet and ~1.5gbps over WIFI.
a m2 ssd eats 2-3watts when writing/reading and less than 1w when idle, whereas a mechanical drive that i have and have seen in general have a dedicated psu of 1.5+Amps or 18w+ i plan to keep the drive online all the time and seed torrents from it. for my use case its the ssd is the more efficient solution power and performance wise. and no i do not see any advantage in having 2 always on devices nas + router powered all the time when i can get a beefy router like ax86u that will do a decent job.

FYI: Using an old mechanical HDD attached to my RT-AX86U my Samba speeds are, Write: 112MB/s (i.e. maximum gigabit) and Read: 86MB/s.

BTW I couldn't find anything to suggest your Orico M2L2-NV03C3-GY is a "rack". It appears to be just a normal SDD to USB enclosure.
congrats your samba server works fine, merlin firmware version ? can pm me or post here your samba config file (/etc/smb.conf) ?
the term rack is used very commonly in my country as a synonym for enclosure, i do realise that there is big diff.
 
congrats your samba server works fine, merlin firmware version ? can pm me or post here your samba config file (/etc/smb.conf) ?
I'm just using the standard config that comes with Merlin 386.5. Performance was no different with 386.4. I also found no noticeable difference between using SMBv1 and SMBv2.

It looks like SMB reading is limited by being single-threaded which is why the CPU load is topping out at exactly 25% (1 of 4 cores). But I haven't seriously looked into that so YMMV.

Side note: See my post here.
 
I haven't done this, and what I'm linking to is an old thread, but could the older version of Samba cause performance issues on some computers? I'm sure a lot of things changed since 23 Feb 2015. Samba is currently on version 4.16.0 and entware has version 4.14.7 available. I don't know the implications of upgrading it with entware or the extra hassle it might add; however, if you want to test if the version is the problem and you have the time, the thread below might give you a head start on testing it.

EDIT: It looks like I should give it a try at some point because Samba 4.8 introduced Time Machine support. It might work better than my current hack that stopped working when I upgraded to MacOS 12.3.

 
Last edited:

Similar threads

Latest threads

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top