[PASS] BanaNAS PRO

44 14900
cyryllo  
Edited by cyryllo at Tue Mar 10, 2015 04:42

Name BanaNAS PRO
Open-source Project Yes
Distribution License AGPLv3 (System)/CC BY-SA 3.0 (case)
The application number 2 piece Pro ver.
Website bananas.gda.pl
Applicant self- introduction I wanted to build a ready case for Banana Pro (3D design a'la Mac mini case) and upgrade a image system of OMV. The project will carry the name "BanaNAS PRO"
The open source content and the first distribute website
Others Project will consist of:
  • ready image with OMV.
  • design your own 3D print in .stl format.


Case will have a place on the BPi and 2.5 "drive. In addition hole For Wifi antena.


Plans for the future:
  • Upgrading packages [done]
  • Kernel update [done]
  • Enable Wifi in PRO [done]
  • Enable multipler Sata (JMB321) [done]
  • Adding Webapplication to configure wifi [future]
  • Preparing the update script (bpi-omv) [to-do]


New img for Banana PRO can download from http://bananas.gda.pl/pobierz/
Single Polls, Total 18 Users voted
100.00% (18)
0.00% (0)
Your User group have no permission to vote
eraz  
Hello Eryk here

Update:
We have succesfully tested 1-to-5 SATA2 Port Multiplier on Banana Pi board. So there is an option to create a RAID version of BanaNAS with built in wifi.
My goal is to develop a casing with:
- 4 bays for 2.5' disk
- banana pi pro board
- 5A 5V power supply
- SATA2 Port Multiplier
- external port for detachable wifi antena
- buttons for power and restart

tkaiser  
Edited by tkaiser at Thu Nov 20, 2014 06:59
eraz replied at Thu Nov 20, 2014 05:09
create a RAID version of BanaNAS


Good luck with building the 'slowest RAID in the world'

If I would plan to build a slow passively cooled RAID based NAS I would go with something like ASRock's Q1900-ITX featuring both the necessary count of SATA connectors as well as the necessary CPU power to do the job. Given the A20's SATA speed limitation and the massive port multiplier slow-down on top (due to Command-based switching mode which affects performance horribly when multiple disks should be accessed simultaneously) the whole idea is somewhat insane

BTW: I would have a look in the specs of the disks you would like to use for their peak current (spin-up). Most of the 2.5" disks that are ready for 24/7 operation require at least 1A peak current.

cyryllo  
For home use the speed enough. For someone who cares about safety and not speed will be ok.

tkaiser  
Edited by tkaiser at Thu Nov 20, 2014 09:02
cyryllo replied at Thu Nov 20, 2014 07:48
For home use the speed enough. For someone who cares about safety and not speed will be ok.


RAID is not about safety but about availability ("business continuity" -- which is also related to realiability but not to safety). And why should I rely on unreliable hardware for this purpose?

A port multiplier that only implements command-based switching will decrease performance to probably 25%-40% compared to single link SATA speed when all disks should be accessed at the same time which will happen when using RAID (remember: this is SATA and not SAS/FC with 'pro' topologies in mind). Please report back when you finished such a setup and report 'client to server real-world throughputs' measured with eg. Helios' LanTest. And also measure I/O performance of the RAID itself both covering sequential as well as random I/O (must be horribly slow).

If you think about 'safety' then it would be better to implement an easy to use syncing solution between two NAS installations that provides an intuitive user interface to restore lost/damaged items from the backup NAS to the productive one. But trying to address 'safety' with an unreliable RAID setup is just fooling yourself

And if you also think about data integrity then the whole approach is even more insane. Using el cheapo desktop/notebook disks with intolerable UER (unrecoverable error rates) together with el cheapo PMs and el cheapo 'server hardware' and then just relying on single redundancy is simply asking for trouble.

RAID sounds like a good idea only as long as you don't identify every SPoF (single point of failure -- and there are so many in your setup compared to the count of disks!) and believe disks die always the 'boom --> dead' way (which is not what they do. They start to fail silently and you will notice this just in the moment when the first disk completely failed -- or the 'controller' thinks the disk is unavailable due to the PM going nuts for a minute -- and the rebuild fails completely due to unrecoverable errors/sectors of other disks... maybe due to the PM going nuts for a minute )

Never ever use home/SOHO RAID unless you internalised the TAO of backup and set up a realiable backup solution implementing versioning and different storage locations. And test full and partial restores from time to time

cyryllo  
tkaiser replied at Thu Nov 20, 2014 08:48
RAID is not about safety but about availability ("business continuity" -- which is also related to ...

Yes, you're right, I did not mean to end data security but about their availability. Of course you can build a really good NAS Raid on the board as provided.

But two disks (or 4) is not what 1 in BanaNAS ;)

tkaiser  
Edited by tkaiser at Fri Nov 21, 2014 09:00
cyryllo replied at Fri Nov 21, 2014 06:23
Of course you can build a really good NAS Raid on the board as provided


Unfortunately this is not true. When you deal with TBytes of storage 'good old' RAID-5 isn't sufficient any longer due to just single redundancy and the statistical failure rates that put your whole array at risk! It's a simple fact. So you need at least double redundancy and that means RAID-6 or similar. And then just 4 drives do not make that much sense.

Using an el cheapo port multiplier as one central single point of failure between host controller and disks does not only mean horribly slow performance in general but also putting your data at even higher risk: 1) if a disk fails -- or the RAID implementation thinks so -- and a rebuild is started and now the PM does something wrong all data is gone instantly. And 2) due to the bad performance the rebuild will take ages which means that meanwhile no other disk is allowed to fail otherwise all data is gone. This alone thwarts the whole RAID approach completely.

If availability / business continuity is a concern (which it isn't at home!) then you have to do it right. And then you should also think about data integrity and that means using a modern file system that implements end-to-end data integrity utilizing checksums over the whole data paths. Which is pointless on 'tablet grade' hardware that lacks at least simple ECC memory.

So the minimum hardware requirements to do RAID correctly if you want to enhance availability and want to care about data integrity (and also want low power consumption, less noise and so on) is at least something like Gigabyte GA-9SISL, SuperMicro A1SAi-2750F or when many disks should be used ASRock C2750D4I or SuperMicro A1SA7-2750F. These octacore avotons with support for large amounts of ECC RAM are fast enough for a lowend RAID. And then you would use eg. FreeBSD or Solaris with ZFS/raidz2/raidz3 to build a reliable RAID implementation featuring also data integrity and that will not fail in 'worst case' scenarios.

And nothing of this addresses even remotely 'data safety'. This can only be done using a separate storage implementation using versioning (no direct mirroring!). RAID won't help here at all.

cyryllo  
Next week we begin to be designed by a new case. It will be a modular enclosure consisting of 2 separate casing.  Something like a new computer Lenovo
case.jpg

External power supply will be.


More information coming soon

cyryllo  
Edited by cyryllo at Wed Jan 28, 2015 05:42

Today we started work on a new version of the case and update the system.

IMG_20150114_122941656.jpg

By the end of the week, we should update the image of the system to handle also the PRO version.

Cubytus  
What speed have you tested with hdparm and a single hard drive? I remember I got a less-than-stellar 60-something on the Bpi, first revision.

You have to log in before you can reply Login | Sign Up

Points Rules