Hello there,
I have a router running OpenWrt 18.06.1 on it, model GL.iNet B1300. I attached a USB 3.0 hard drive to the router. When I download file from the USB hard drive, the speed is only about 3 MB/s. That is much less than my expectation.
So, what should the normal speed for this be?
P.S., I SSH to the router, and tested R/W speed of the USB hard drive with dd command. The result is about 38 MB/s for writing and 80 MB/s for reading.
SSH, NFS, SMB, AFP…? which one are u sing for the file transfer? Can you check the speed difference between SSH file transfer and SMB?
If you are on MacOS you will get better speed with AFP and NFS than SMB, Even if you are on Windows or Linux you will get less transfer speed with SMB than the Mac with AFP.
Its an SMB issue not the router
Though I dont think the device is powerfull enough to sustain the NAS function. Only occasionnal share
I am on macOS and tried SFTP and AFP, both are about 2~3 MB/s download speed.
This is too slow, AFAICT. BTW, you just can't justify from this kind of test.
However, you can try to setup NFS and use it to test. In my case with a Seagate GoFLEX Home running on a debian Linux with NFS, I get an average of 70/30 MBps (R/W) speeds. Below is the result of some plain (no encryptions) R/W speed tests from a debian Linux computer to my Seagate GoFLEX Home using F3 utility from Digariti.
[mazilo@linux:/home/local/PEOPLE/mazilo 388%] ~ f3write /mnt/devel/junk/
Free space: 2.44 TB
Creating file 1.h2w ... OK!
Creating file 2.h2w ... 0.05% -- 30.10 MB/s -- 23:38:14^C
2.001u+3.763s=0:40.18e(14.3%) TDSavg=0k+0k+0k max=2244k 80+2467200io 0pf+0sw
[mazilo@linux:/home/local/PEOPLE/mazilo 389%] ~ f3read /mnt/devel/junk/
SECTORS ok/corrupted/changed/overwritten
Validating file 1.h2w ... 2097152/ 0/ 0/ 0
Validating file 2.h2w ... 369712/ 0/ 0/ 0
Data OK: 1.18 GB (2466864 sectors)
Data LOST: 0.00 Byte (0 sectors)
Corrupted: 0.00 Byte (0 sectors)
Slightly changed: 0.00 Byte (0 sectors)
Overwritten: 0.00 Byte (0 sectors)
Average reading speed: 69.73 MB/s
2.715u+1.222s=0:17.36e(22.6%) TDSavg=0k+0k+0k max=2244k 2466928+0io 0pf+0sw
[mazilo@linux:/home/local/PEOPLE/mazilo 390%] ~
Also, install iperf3 on your NAS and/or the computer you want to test. Then, you can do some R/W tests as follows:
From the above test, it will show you the network bandwidth throughputs you can achieve from your computer to your NAS and/or vice versa. The following shows the output from an iperf3 test from my debian Linux computer to my Seagate GoFLEX Home.
[root@linux:/home/local/PEOPLE/mazilo 391%] # iperf3 -c 192.168.1.88
Connecting to host 192.168.1.88, port 5201
[ 4] local 192.168.1.74 port 43818 connected to 192.168.1.88 port 5201
[ ID] Interval Transfer Bandwidth Retr Cwnd
[ 4] 0.00-1.00 sec 72.8 MBytes 610 Mbits/sec 0 365 KBytes
[ 4] 1.00-2.00 sec 92.3 MBytes 775 Mbits/sec 0 399 KBytes
[ 4] 2.00-3.00 sec 90.1 MBytes 756 Mbits/sec 0 423 KBytes
[ 4] 3.00-4.00 sec 90.2 MBytes 757 Mbits/sec 0 423 KBytes
[ 4] 4.00-5.00 sec 90.3 MBytes 758 Mbits/sec 0 423 KBytes
[ 4] 5.00-6.00 sec 89.8 MBytes 753 Mbits/sec 0 423 KBytes
[ 4] 6.00-7.00 sec 91.2 MBytes 766 Mbits/sec 0 423 KBytes
[ 4] 7.00-8.00 sec 91.4 MBytes 767 Mbits/sec 0 423 KBytes
[ 4] 8.00-9.00 sec 90.5 MBytes 759 Mbits/sec 0 423 KBytes
[ 4] 9.00-10.00 sec 89.7 MBytes 753 Mbits/sec 0 423 KBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 888 MBytes 745 Mbits/sec 0 sender
[ 4] 0.00-10.00 sec 887 MBytes 744 Mbits/sec receiver
iperf Done.
0.032u+1.621s=0:10.18e(16.2%) TDSavg=0k+0k+0k max=2412k 0+0io 1pf+0sw
[root@linux:/home/local/PEOPLE/mazilo 392%] #
As you can see from the above, I can only achieve 10% in read tests from the network bandwidth read throughput. Ideally, I would love to be able to get at least a double in R/W (particularly the later) speed tests, respectively. Unfortunately, as you can see the bottleneck will become the gigabit network traffics. I am hoping some experts here can help.
BTW, a hdparm -tT /dev/sda test (see below) on my Seagate GoFLEX Home indicates the HDD on my Seagate GoFLEX Home is capable of delivering more R/W throughputs.
[root@debian:/root 2%] # hdparm -tT /dev/sda
/dev/sda:
Timing cached reads: 522 MB in 2.01 seconds = 260.16 MB/sec
Timing buffered disk reads: 382 MB in 3.01 seconds = 126.72 MB/sec
0.302u+4.301s=0:13.57e(33.8%) TDSavg=0k+0k+0k max=3896k 798816+0io 5pf+0sw
[root@debian:/root 3%] #
I meant 60%.
Can you run top -d 5 to see if an I/O process takes up CPU and/or memory?
Perhaps there's an I/O bottleneck.
For a 717 MHz single core CPU - running at 1000 Mbps and full USB 3.0 transfer speed...
Normal speed might be: Slow.
It should be a quad-core CPU. Some (on other forums) says that 4~6 MB/s is the normal speed.
I stand corrected, it does say quad core.
What does top show?
This is the output of top, with NFS upload and download running
Mem: 224336K used, 26608K free, 136K shrd, 5820K buff, 149212K cached
CPU: 13% usr 9% sys 0% nic 47% idle 14% io 0% irq 14% sirq
Load average: 2.38 2.66 1.91 2/106 2718
PID PPID USER STAT VSZ %VSZ %CPU COMMAND
2007 1 root S 14648 6% 19% transmission-daemon -g /mnt/transmission -f
1794 2 root SW 0 0% 2% [nfsd]
1793 2 root SW 0 0% 1% [nfsd]
182 2 root SW 0 0% 1% [usb-storage]
7 2 root RW 0 0% 0% [ksoftirqd/0]
33 2 root SW 0 0% 0% [kswapd0]
88 2 root IW 0 0% 0% [kworker/3:2]
1799 1 root S 2556 1% 0% /usr/sbin/rpc.mountd -p 32780 -F
2648 2626 root R 1080 0% 0% top -d 5
1792 2 root SW 0 0% 0% [nfsd]
1494 1 root S 1628 1% 0% /usr/sbin/hostapd -s -P /var/run/wifi-phy0.pid -B /va
5 2 root IW 0 0% 0% [kworker/u8:0]
84 2 root IW 0 0% 0% [kworker/1:1]
231 2 root SW 0 0% 0% [jbd2/sda2-8]
14 2 root SW 0 0% 0% [ksoftirqd/1]
24 2 root SW 0 0% 0% [ksoftirqd/3]
754 1 root S 1528 1% 0% /sbin/netifd
2625 1007 root S 892 0% 0% /usr/sbin/dropbear -F -P /var/run/dropbear.1.pid -p 2
8 2 root IW 0 0% 0% [rcu_sched]
27 2 root IW 0 0% 0% [kworker/0:1]
19 2 root SW 0 0% 0% [ksoftirqd/2]
Seems nothing wrong
try stopping transmission
From the output of your top utility, it only shows about 3 nfsd threads. Can you please post the output from nfsstat -m
io is quite highsirq (Software IRQs) is at 14%...again, if you're idle, this is high. This clearly indicates to me a bottleneck!I agree with @fuller:
I tested with iperf3, transfer speed from my laptop to the router is only 2.4 MB/s.
And the result of hdparm -tT /dev/sda is
/dev/sda:
Timing cached reads: 692 MB in 2.00 seconds = 345.54 MB/sec
Timing buffered disk reads: 272 MB in 3.03 seconds = 89.92 MB/sec
Sorry, but which package contains the nfsstat command? I can't find it.
I tried to stop transmission. The performance of iperf3 raised from 2.4 to 2.7 MB/s...
I've heard that transmission is performance hungry, so is it normal to take over 20% of CPU when it downloading at about 5 MB/s speed?
BTW, the transmission-web client gets very slow when transmission is downloading, but this doesn't happen when transmission is only uploading.
and, was it successful?
yes, especially with parallel torrents it tends to make random writes to hdd causing it to react very slow.
hence it would have perfectly explained your situation... 
No. Stopping transmission just make 0.3 MB/s increase in NAS speed
You exclude a lot of essential information so let's try troubleshooting 101
Transmission usually pre-allocates files which puts a high load on I/O and during that time the UI will feel sluggish, this also happens on beefier devices so it's normal as far as Transmission goes. This is a design "limitation" by transmission, I guess the UI polls the I/O part of transmission and waits instead of dropping the request which makes it feel unresponsive.
I'd say that expected transfer rates should be around 35-50mbyte/s using Samba4 and USB3 (ext3/4), NFS should be slightly faster and using filesystems such as btrfs is going to slow down things quite a bit.
Keep in mind that IPQ series should still be considered "experimental" SoCs as non are supported in mainline so you might run into platform quirks. I'm not sure how much the support has improved in 4.19 but a few things like USB probably works quite a bit more efficient.
Iperf3 directly tests the link, you can never get filesystem performance faster than this link speed. I assume this is wifi. Test your filesystem issue over wired connection, or if this is wired, try different network cables!!!!
If this is wifi try changing channels to 1,6,or 11 on 2.4ghz and to a non DFS channel that isn't in use near you on 5ghz.
That's a pretty low throughput unless the test was done through a WiFi connection. There seems to be some bottleneck somewhere between your computer and your NAS. If you have the F3 utility installed on your NAS, give it a try to see what is the R/W throughput you can get from your NAS to its HDD. See below the result of the R/W tests I did using F3 utility on my NAS.
[debian@debian:/home/local/PEOPLE/debian 5%] ~ f3write /srv/devel/junk/
F3 write 7.1
Copyright (C) 2010 Digirati Internet LTDA.
This is free software; see the source for copying conditions.
Free space: 2.44 TB
Creating file 1.h2w ... OK!
Creating file 2.h2w ... 0.07% -- 50.00 MB/s -- 14:11:35^C
5.021u+23.346s=0:38.18e(74.2%) TDSavg=0k+0k+0k max=3604k 856+3633672io 3pf+0sw
[debian@debian:/home/local/PEOPLE/debian 6%] ~ f3read /srv/devel/junk/
F3 read 7.1
Copyright (C) 2010 Digirati Internet LTDA.
This is free software; see the source for copying conditions.
SECTORS ok/corrupted/changed/overwritten
Validating file 1.h2w ... 0/ 2097152/ 0/ 0
Validating file 2.h2w ... 0/ 1536156/ 0/ 0
Data OK: 0.00 Byte (0 sectors)
Data LOST: 1.73 GB (3633308 sectors)
Corrupted: 1.73 GB (3633308 sectors)
Slightly changed: 0.00 Byte (0 sectors)
Overwritten: 0.00 Byte (0 sectors)
Average reading speed: 82.43 MB/s
4.182u+14.698s=0:21.58e(87.4%) TDSavg=0k+0k+0k max=1656k 3633480+0io 4pf+0sw
[debian@debian:/home/local/PEOPLE/debian 7%] ~
On my Seagate GoFLEX Home running debian OS, dpkg -S /usr/sbin/nfsstat indicates the nfsstat utility is part of nfs-common package.