GnuBee Personal Cloud 2 -- 'Open Source NAS' for 6 x 3.5" disks

Quoting the campaign on CrowdSupply: A low-cost, low-power NAS for 3.5" drive

This is some sort of an open enclosure combined with a PCB based on a MIPS MediaTek SoC with 3 GbE ports and 3 ASM106x SATA controllers hanging off a dedicated PCIe lane each so in theory a pretty decent setup to access up to 6 SATA disks. Unfortunately I've not seen any reasonable benchmark numbers so far using this SoC accessing more than one disk in parallel (all I've came accross were single disk benchmarks with slow 2.5" HDD or these pretty low numbers from Mqmaker's WiTi board -- but as we know without some tweaks slow ARM or MIPS SoCs won't perform well as OMV/NAS boxes anyway).

Some thoughts/considerations as usual on CNX: GB-PC2 as well as the older 2.5" variant (check especially the comments section)

While I really love the idea behind (creating a NAS relying only on FLOSS components) I don't understand the implementation (obviously only outdated kernels available, loading a binary firmware recommended and since this device needs Debian mipsel architecture you're dependent on such a repo for OMV too). Since GNUBee folks advertise OMV capabilities they might comment here?

Anyway: it feels kinda bizarre assuming that a MIPS based NAS thingie couldn't be the right choice in 2017 given that almost all the customers 2 decades ago that had really large storage setups were running on MIPS (SGI servers with IRIX and XFS which wasn't available anywhere else back then) and that the most beautiful UNIX box I ever had running also felt pretty fast wrt storage back then

A friend more into the hardware part of things mentioned only 'Good luck with vibrations' when looking at the picture. I've to admit that I realized only after his remark that spindle centers are on top and even a few cm above the metal plates.

Zitat von ness1602

looks good

Hmm... after thinking a little bit longer about center of gravity, rotation, vibrations, impact on performance (maybe longevity too?) the device with its current design (the side plates) definitely does not look good to me any longer

And then dealing with a Debian mipsel architecture here is something different compared to the officially supported architectures (speaking about armhf and arm64)

Zitat von ryecoaaron

Why would this mounting be any different than a server that stores the spindles vertically (like an N40L) that doesn't have any vibration dampening (stored rigidly in a hot swap tray)?

Since it's somewhat different if you only fix the drive at 4 points below gravity centre with a the spindles allowed to swing freely above?

I did a quick test with a 3TB Seagate Barracuda one time fixed in a drive cage, the other time simulating the GnuBee fixation using 'iozone -e -I -a -s 100M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2':

GnuBee style                                                  random    random
              kB  reclen    write  rewrite    read    reread    read     write
          102400       4    58791    73148    69835    70883     1152      919
          102400      16   112256   149872   138120   146112     4325     3830
          102400     512   148584   148143   154176   163686    60286    99389
          102400    1024   146546   144482   156688   166931    85745   115663
          102400   16384   132835   148271   158633   161372   152230   132610


Fixed                                                         random    random
              kB  reclen    write  rewrite    read    reread    read     write
          102400       4    64330    71447    69835    72784     1153      924
          102400      16   135730   153744   143659   152118     4350     4311
          102400     512   150105   149943   154109   163434    61183    95582
          102400    1024   148360   148200   154675   163333    86938    83239
          102400   16384   138317   151907   151418   161071   154072   136433

Only minimal advantages with 16K blocksize but this might change once there are a few more disks adding to overall vibrations? I wonder whether GnuBee folks should not consider making the side plates taller and fix the disks on top too.

Edit: One of my personal heroes explaining how they found one missing screw not properly fixating a disk somewhere in a huge storage box with dtrace since this single disk showed latency too high. Worth watching IMO

Maybe good news (see the campaign update

Most probably bad news: https://github.com/gnubee-git/…31#issuecomment-341720837

(I still don't understand the whole 'libre software' advertisements when users depend on vendor repos for their software and can't build stuff themselves)