Entware Entware: which USB stick performs well?

kvic · Sep 6, 2018

sfx2000 said:
It's not one of these is it?

Guess?

sfx2000 said:
The armbian team has made the most progress on the rockchip series of SoC's -- lot of good insight shared in their forums.

My first and brief impression about Armbian was very good. I'm tired of Debian and I don't like Ubuntu. As a result, I can't stay with them.

ATM, my ALARM on 4.4.154 is pretty good... efficiency, performance wise. Main line 4.18.6 could run as well. I'm lacking SDRAM timing data and hence it's no fun to run the mainline kernel performance wise.

sfx2000 · Sep 6, 2018

kvic said:
Guess?

Nice - one of the folks to really follow over there is TonyMac32 - he's one the smarter guys on the RK chips, and has done a lot of good work there.

kvic · Sep 7, 2018

sfx2000 said:
Nice - one of the folks to really follow over there is TonyMac32 - he's one the smarter guys on the RK chips, and has done a lot of good work there.

I spent some time there but not much. Perhaps I shall open an account to post but then I don't want to end up in a situation like here. lol.

I like his posts about heatsink. That brought me to Google and led me to some fascinating articles about heatsink. For the first time of my life, I learned diamond is a superb good heat conductor.

sfx2000 · Sep 15, 2018

sfx2000 said:
Or any of the Samsung portable SSD's...

just sharing... N3700 NUC with Samsung 850 EVO...

Code:

Command line used: ./iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2 /home/sfx
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
                                                            random  random    bkwd   record   stride                                  
              KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite frewrite   fread  freread
          20480       4   50615   66116    70991    67584   28551   61632                                                         
          20480      16  135983  156652   165272   163647   95163  145680                                                         
          20480     512  375531  383842   396001   403030  391707  366912                                                         
          20480    1024  356297  401026   429205   439268  403211  380775                                                         
          20480   16384  221023  430228   493508   512526  524320  454113                                                         
iozone test complete

sfx2000 · Sep 16, 2018

Here's an older Samsung EVO 32GB SD-Card...

(it's not a EVOPlus - it's really old) - captured on Asus Tinkerboard running Armbian...

Review on the same card here -- Storage Review

And yes, the card supports TRIM...

Code:

lsblk --discard
NAME        DISC-ALN DISC-GRAN DISC-MAX DISC-ZERO
mmcblk0            0        4M     3.5G         0
└─mmcblk0p1        0        4M     3.5G         0

Anyways - here's the IOZone stats - not bad for an old card - still has some legs there... I've beat this one up a lot across different development projects...

Code:

    Command line used: ./iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2 /home/sfx
    Output is in Kbytes/sec
    Time Resolution = 0.000001 seconds.
    Processor cache size set to 1024 Kbytes.
    Processor cache line size set to 32 bytes.
    File stride size set to 17 * record size.
                                                            random  random    bkwd   record   stride                                   
              KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite frewrite   fread  freread
           20480       4    1167    1263    12286    12253   12712    4606                                                         
           20480      16    9977   10053    25712    24181   24476   10252                                                         
           20480     512   29964   30400    42978    42318   40999   30196                                                         
           20480    1024   29894   30308    44544    43451   41741   29940                                                         
           20480   16384   27384   30769    46268    46277   46237   30636                                                         

iozone test complete.

kvic · Sep 24, 2018

sfx2000 said:
Here's an older Samsung EVO 32GB SD-Card...

(it's not a EVOPlus - it's really old) - captured on Asus Tinkerboard running Armbian...

Review on the same card here -- Storage Review

And yes, the card supports TRIM...

Very good iozone result! This is from the previous generation(s) of Samsung flash chips - the yellow EVO and the red EVO+. They also perform much better than the latest generation - the red EVO Plus...not "+". Confusing enough for consumers.

The few flash stick/microsd cards mentioned recently were all added into iozone-db.txt. With this yellow EVO done today. Sorry for the latency

codeone · Sep 24, 2018

Anyone feel like they can explain the 4k and 16k write/rewrite results below?
(Also not sure why my time on the "Run began" is off?)

Some details:
Using a SanDisk Ultra CZ48 32GB USB 3.0 in the USB 2 port (can't get it to mount or work at all in the USB 3 port.
2 partitions, (4GB and 28GB) both formatted as EXT2.

Was using a SanDisk SDCZ60 16GB Cruzer USB 2.0 in the USB 3 port that got 4k and 16k write/rewrite numbers about double these...

Code:

    Run began: Mon Sep 24 14:04:34 2018

    Include fsync in write timing
    O_DIRECT feature enabled
    Auto Mode
    File size set to 20480 kB
    Record Size 4 kB
    Record Size 16 kB
    Record Size 512 kB
    Record Size 1024 kB
    Record Size 16384 kB
    Command line used: /jffs/bin/iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
    Output is in kBytes/sec
    Time Resolution = 0.000001 seconds.
    Processor cache size set to 1024 kBytes.
    Processor cache line size set to 32 bytes.
    File stride size set to 17 * record size.
                                                              random    random     bkwd    record    stride                                  
              kB  reclen    write  rewrite    read    reread    read     write     read   rewrite      read   fwrite frewrite    fread  freread
           20480       4      488      510     7514     7494     4125      525                                                              
           20480      16     1885     1820    17950    18646    13260     1848                                                              
           20480     512    10830    14853    28350    27657    29126    14618                                                              
           20480    1024    14685    15512    29435    30528    30279    16187                                                              
           20480   16384    14625    17436    31444    31935    32258    19095

Not noticing any problems just curious.

Anyone have advice on getting this to mount in the USB 3 port?

kvic · Sep 24, 2018

codeone said:
Using a SanDisk Ultra CZ48 32GB USB 3.0 in the USB 2 port (can't get it to mount or work at all in the USB 3 port.

Most likely a combination of USB host controller + driver in kernel issue. Have you tried: unplug power cord, plug in flash drive and then plug in power cord?

Btw the data of this stick looks like new to my db. Thankyou.

Was using a SanDisk SDCZ60 16GB Cruzer USB 2.0 in the USB 3 port that got 4k and 16k write/rewrite numbers about double these...

Another member got this stick and good result too...and he got it at deep discount!

M@rco · Oct 2, 2018

As my Sandisk Ultra Flair 32 GB USB 3.0 drive refuses to get recognized as USB 3.0 drive, I have been running it as a USB 2.0 device and it's rather slow. I found a couple of Kingston DataTraveler SE9 64 GB USB 3.0 drives, formatted them and they get recognized as a USB 3.0 drive instantly.
I benchmarked one of them, using ext 2 and ext 4 with journalling enabled, mounted as USB 3.0 drive, with iozone. Results are below.

Formatted as ext2:

Code:

        Run began: Tue Oct  2 21:17:45 2018

        Include fsync in write timing
        O_DIRECT feature enabled
        Auto Mode
        File size set to 20480 kB
        Record Size 4 kB
        Record Size 16 kB
        Record Size 512 kB
        Record Size 1024 kB
        Record Size 16384 kB
        Command line used: /tmp/mnt/sandisk/iozone/iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
        Output is in kBytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 kBytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                              random    random     bkwd    record    stride
              kB  reclen    write  rewrite    read    reread    read     write     read   rewrite      read   fwrite frewrite    fread  freread
           20480       4     1548     1550     6767     6782     6818     1313
           20480      16     4920     4098    21962    22104    21111     3960
           20480     512    12850    12617    72584    71326    72811    39423
           20480    1024    12386    12739    67780    75286    75376    39546
           20480   16384    10697    11884    74678    70793    75376    41421

iozone test complete.

Formatted as ext4 with journalling enabled:

Code:

Run began: Tue Oct  2 21:23:52 2018

        Include fsync in write timing
        O_DIRECT feature enabled
        Auto Mode
        File size set to 20480 kB
        Record Size 4 kB
        Record Size 16 kB
        Record Size 512 kB
        Record Size 1024 kB
        Record Size 16384 kB
        Command line used: /tmp/mnt/sandisk/iozone/iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
        Output is in kBytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 kBytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                              random    random     bkwd    record    stride
              kB  reclen    write  rewrite    read    reread    read     write     read   rewrite      read   fwrite frewrite    fread  freread
           20480       4     1568     1600     6850     7039     6942     1428
           20480      16     6136     6248    22077    23324    23442     6196
           20480     512    14480    23548    84334    84895    79860    65998
           20480    1024    23683    65408    79618    86113    83472    23987
           20480   16384    21061    68921    88947    90748    90540    69899

iozone test complete.

While benchmarking this thumb drive was entirely empty, so no entware installed yet, and no read/write actions on the drive besides the activity by iozone.

kvic · Oct 3, 2018

M@rco said:
As my Sandisk Ultra Flair 32 GB USB 3.0 drive refuses to get recognized as USB 3.0 drive, I have been running it as a USB 2.0 device and it's rather slow. I found a couple of Kingston DataTraveler SE9 64 GB USB 3.0 drives, formatted them and they get recognized as a USB 3.0 drive instantly.

As mentioned above, it's most likely due to USB 3 driver issues. For running apps, you should look at 4K random read/write. USB 2 vs 3 makes little difference. So your Ultra Flair should not be slow in this regard.

kvic · Oct 8, 2018

The article "Benchmark usb flash drives on routers" referred to in post #1 is now accessible. Restored from backup and now hosted on a new VPS.

loveleeyoungae · Nov 18, 2018

My "My Passport Ultra" has good other read/write speed, but the 4k random read/write is extremely low. Would that affect apps? I've just installed Diversion and Skynet, felt a bit glitch here and there.

Code:

        Command line used: /opt/bin/iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
        Output is in kBytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 kBytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                              random    random     bkwd    record    stride
              kB  reclen    write  rewrite    read    reread    read     write     read   rewrite      read   fwrite frewrite    fread  freread
           20480       4    14255    19511    17728    20229      474     2250
           20480      16    38103    49357    39897    42124     2102    11740
           20480     512    76557    93979    79364    80100    35338    78503
           20480    1024    73159    89473    65078    82612    50200    93827
           20480   16384    92338    91403    86784    86532    87063   100000

iozone test complete.

kvic · Nov 18, 2018

loveleeyoungae said:
felt a bit glitch here and there.

Then just get a faster stick. It helps..sometimes noticeably on demanding apps such as syslog-ng. Note that people reported Merlin FWs compatibility with USB 3 is a bit "hit and miss." You might not be able to take full advantage of a USB3 stick. Regardless it should give you a better 4K experience even on USB 2.

loveleeyoungae · Nov 18, 2018

kvic said:
Then just get a faster stick. It helps..sometimes noticeably on demanding apps such as syslog-ng. Note that people reported Merlin FWs compatibility with USB 3 is a bit "hit and miss." You might not be able to take full advantage of a USB3 stick. Regardless it should give you a better 4K experience even on USB 2.

Actually, the My Passport Ultra is a HDD-type drive by Western. It is really fast, I never suspected it has such low 4k random read/write.

And yeah, regarding USB 3.0 issue: this HDD USB drive survives after power cycle/reboot, always mounted and performs good speed for normal file transfer. However, my 16GB Sandisk Flair (I bought per your test) can't survive after power cycle/reboot on USB 3.0 - the only way for the router to recognize it is unplug it, power cycle (NOT reboot) the router, then plug it. Change the usb mode to USB 2.0 is the only way to have it always mounted.

Thank you for your info.

kvic · Nov 18, 2018

loveleeyoungae said:
my 16GB Sandisk Flair (I bought per your test) can't survive after power cycle/reboot on USB 3.0 - the only way for the router to recognize it is unplug it, power cycle (NOT reboot) the router, then plug it.

For those who bought this stick on my recommendation, and only realise later that doesn't work quite as compatible on Merlin FWs, I'm sorry..

My RT-AC56U running 380.66 has no such compatibility issue. So perhaps safe to say that it only happens after that and in more recent FWs.

BTW, your observation on issue wrt power-up/reboot is spot on I've to say! I've seen exact symptom in other places. Fixable in USB drivers.

cmkelley · Nov 18, 2018

SanDisk Ultra USB 3.0 16GB on AC86U USB3.0 port, formatted ext4, no journaling.

Code:

        Run began: Sun Nov 18 18:47:23 2018

        Include fsync in write timing
        O_DIRECT feature enabled
        Auto Mode
        File size set to 20480 kB
        Record Size 4 kB
        Record Size 16 kB
        Record Size 512 kB
        Record Size 1024 kB
        Record Size 16384 kB
        Command line used: /jffs/bin/iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
        Output is in kBytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 kBytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                              random    random     bkwd    record    stride
              kB  reclen    write  rewrite    read    reread    read     write     read   rewrite      read   fwrite frewrite    fread  freread
           20480       4     2341     2343    18049    18151     6450      849
           20480      16    15202    12748    50521    50610    18955     1312
           20480     512    75259    76249   130280   130869   103197     7583
           20480    1024    31909    75440   129081   130384   106663    19553
           20480   16384    34349    69377   130674   131946   131985    69602

iozone test complete.

cmkelley · Nov 28, 2018

I happened to have an "old" (2015) 120GB Samsung 850 EVO SSD installed in a laptop I primarily use for fiddling around with different operating systems. So I bought one of these: https://www.amazon.com/gp/product/B06XWRRMYX/?tag=snbforums-20 for $10, popped the SSD out of the laptop, connected it to my router, and ran the iozone test for fun.

This is the 120GB Samsung 850 EVO SSD in a UGREEN External Hard Drive Enclosure Adapter, on my AC86U USB 3.0 port, formatted ext4, with journaling:

Code:

 Run began: Tue Nov 27 21:16:33 2018

        Include fsync in write timing
        O_DIRECT feature enabled
        Auto Mode
        File size set to 20480 kB
        Record Size 4 kB
        Record Size 16 kB
        Record Size 512 kB
        Record Size 1024 kB
        Record Size 16384 kB
        Command line used: /jffs/bin/iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
        Output is in kBytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 kBytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                              random    random
              kB  reclen    write  rewrite    read    reread    read     write
           20480       4    39567    46394    23520    23579    23420    45564                                                     
           20480      16   114546   129076    64957    64515    64710   126096                                                     
           20480     512   281129   291028   131670   133172   131223   281811                                                     
           20480    1024   290319   296299   132707   130811   133040   295504                                                     
           20480   16384   299251   298742   132731   135134   134524   298313                                                     

iozone test complete.

Same SSD & router, formatted ext4, no journaling:

Code:

       Run began: Tue Nov 27 22:17:55 2018

        Include fsync in write timing
        O_DIRECT feature enabled
        Auto Mode
        File size set to 20480 kB
        Record Size 4 kB
        Record Size 16 kB
        Record Size 512 kB
        Record Size 1024 kB
        Record Size 16384 kB
        Command line used: /jffs/bin/iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
        Output is in kBytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 kBytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                              random    random
              kB  reclen    write  rewrite    read    reread    read     write
           20480       4    40312    43049    21736    21767    21415    43593
           20480      16   111006   121536    60982    60403    59746   119847
           20480     512   296373   271939   125030   130890   123932   270120
           20480    1024   266192   286145   123595   123997   127692   264016
           20480   16384   261587   295836   130066   136989   136970   291126

iozone test complete.

Same SSD & router, formatted ext2:

Code:

        Run began: Tue Nov 27 22:12:32 2018

        Include fsync in write timing
        O_DIRECT feature enabled
        Auto Mode
        File size set to 20480 kB
        Record Size 4 kB
        Record Size 16 kB
        Record Size 512 kB
        Record Size 1024 kB
        Record Size 16384 kB
        Command line used: /jffs/bin/iozone -e -I -a -s 20M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
        Output is in kBytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 kBytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                              random    random
              kB  reclen    write  rewrite    read    reread    read     write
           20480       4    43489    47106    23571    23489    23362    46687
           20480      16   122021   131876    63051    64416    65426   129548
           20480     512   279716   284183   126684   128611   128555   286505
           20480    1024   289600   294015   132112   132991   132601   292902
           20480   16384   301591   294813   136379   136804   136847   302421

iozone test complete.

So, in this specific instance, filesystem and journaling have little effect; less than 4% from worst random write to best. Random read has less than a 2% difference. I don't claim these result are "typical", but it's not a difficult test to do yourself if you have an empty drive to play with.

gattaca · Nov 29, 2018

cmkelley said:
I happened to have an "old" (2015) 120GB
... So I bought one of these: https://www.amazon.com/gp/product/B06XWRRMYX/?tag=snbforums-20 for $10, popped the SSD out of the laptop, connected it to my router, and ran the iozone test for fun.

Yes, I just brought this same UGREEN unit based on the great reviews and the better chipset they used! There's a 3.1 version as well. I plan to do the same thing you did with an older Intel 160MB SSD I have laying around this weekend and then try to load Diversion/Skynet/...

From the iostats, this route seems to be best way for me to get the best I/O I can for the router given I have a couple of these smaller SSD laying around collecting dust. As a performance guy, I/O is always the slowest bottleneck in the system and this seems like a good way to circumvent that issue permanently. The last thing I need is the router misbehaving which leads to me getting calls at work about the internet being down from home!

Thanks!

cmkelley · Nov 29, 2018

gattaca said:
Yes, I just brought this same UGREEN unit based on the great reviews and the better chipset they used! There's a 3.1 version as well. I plan to do the same thing you did with an older Intel 160MB SSD I have laying around this weekend and then try to load Diversion/Skynet/...

I don't think any routers have USB 3.1 Gen 2 (10Gbps) yet so USB 3.0 (which has been confusingly renamed USB 3.1 Gen 1) is fine for use with a router.

gattaca · Nov 30, 2018

cmkelley said:
I don't think any routers have USB 3.1 Gen 2 (10Gbps) yet so USB 3.0 (which has been confusingly renamed USB 3.1 Gen 1) is fine for use with a router.

Correct. I was letting folks know there are both 3.0 and 3.1 versions available. The 3.1 version is about $18 vs $10 for the 3.0. While 3.1 is suppose to be backward compatible, we all know that with USB "compatibility," depends on the chipset and vendor --> and YMMV wildly! Later.

Thread starter	Title	Forum	Replies	Date
J	Low-Power SSD for amtm, Diversion, Entware and Tailscale on the USB 2.0 Port (USB 3.0 configured as USB 2.0)	Asuswrt-Merlin	130	Jan 29, 2024
M	AC68U - RTL2832U and Entware	Asuswrt-Merlin	5	Aug 31, 2024
W	Question: About AX4200 regarding SSH support and Entware	Asuswrt-Merlin	23	Jul 8, 2024
D	How would you go about syncing/symlinking root home folder to entware drive during system operation?	Asuswrt-Merlin	12	Apr 5, 2024
J	Format required for external drive with scripts + entware?	Asuswrt-Merlin	7	Jan 10, 2024
C	Entware doesn't get mounted after reboot on DSL-AC68U	Asuswrt-Merlin	0	Jan 10, 2024
S	No file transfer to USB drive	Asuswrt-Merlin	3	Oct 5, 2024
M	RT-AX86U and DNS Privacy over USB WAN	Asuswrt-Merlin	0	Sep 12, 2024
F	AX86u USB transfer rate	Asuswrt-Merlin	2	Aug 15, 2024
F	Asus USB-C2500 2.5GB Adapter	Asuswrt-Merlin	1	Jul 31, 2024

Entware Entware: which USB stick performs well?

Part of the Furniture

Part of the Furniture

Part of the Furniture

Part of the Furniture

Part of the Furniture

Part of the Furniture

New Around Here

Part of the Furniture

M@rco

Guest

Part of the Furniture

Part of the Furniture

Regular Contributor

Part of the Furniture

Regular Contributor

Part of the Furniture

Very Senior Member

Very Senior Member

Senior Member

Very Senior Member

Senior Member

Similar threads

Similar threads

Support SNBForums w/ Amazon

Sign Up For SNBForums Daily Digest