« I'm back.. or am I in front? | Main | Chenbro ES34069-Mini-ITX Home Server/NAS Chassis as new unRAID server »
May 01, 2008
unRAID unvieled. my new NAS OS.
Today I bring to view, unRAID from LIME TECHNOLOGY.
This has been something I've been exploring for quite some time now. As all the hard drives in my network started filling up from DVD rips I knew there had to be a better way. It's easy enough to buy a new hard drive, but then you have to consider how do you backup the precious data you've just placed on it.. Buy another, then raid them.
The downside of this is, you are always growing and you have to keep buying 2 drives for raid-1 or growing filesystems across multiple drives. Geeze, then god forbid you loose two drives in a raid-5 set up! This happened to me once and I'll never forget it.
I also was considering just how many spindles are spinning on the network for each raid set up. As it grows, so do the costs of keeping them live.
Don't get me wrong here, I love my readyNAS NV+. It's a great lil box, but I noticed I was outgrowing the 4 slots and realized how much time it takes to grow the file systems across multiple spindles.
Granted, RAID5 across multiple spindles has it's advantages,
Redundancy, Fast read speed, Large filesystem support.
But it also has it's cons such as...
Difficulty in expansion, How do you backup a multi terabyte raid 5 filesystem and what happens when you loose more then one spindle in a raid 5 environment.
I've had this issue in the past where I lost 2 spindles (due to simpel IDE lockups) and the raid 5 array was shot... Many mp3's from napster lost.. Oh well.
What I really like about unRAID is that each hard drive spindle becomes it's own filesystem. To me this has the pro of easy migration to larger hard drives, and the con of managing multiple hard drives and navigation across them.
unRAID solves all my neesd for a Media Storage (or backup storage) server.
It's actually a protected JBOD. Similar to RAID4 but without the striping across multiple spindles.
Each Hard drive (of any size) uses a shared parity drive (must be the largest in the system), granted there is a slight performance penalty, but for a media storage box it far outweighs the management of a mult-spindle raid5 arrangement.
Because each drive is it's own filesystem, you can litterally stop the raid array, put in a larger hard drive, restart, and the volume will be rebuilt/expanded onto the larger drive.
In addition since each drive shares the parity, the chances of loosing all your data to a multiple drive failure drop dramatically.
I.E. The array will survive a single drive failure and rebuilt the failed drive when replaced.
If by some chance, you loose a second drive, then you only loose the data on those two drives. The other remaining drives will still be intact because they have thier own separate file systems.
My only beef is that the author users reiserfs instead of ext3.. but it's a small price to pay
On the plus side, The unraid environment is designed to run the root filesystem in ram.
All the data drives are mounted, but programmed to spind down on inactivity.
This became a major plus for me with all the hard drives spinning on my network, and summer coming along with no a/c in my computer area..
Another cool feature of unRAID is it's userspace filesystem.
Because you have multiple drives and data may not all fit on one spindle, there is the user space filesystem that sort of joins the multiple drives into a single view.
Let's say for example you have 10 disks and a DVD folder.
If you program the share to overflow onto 5 of the disks, when one of the drives starts to fill up the user space filesystem will automatically write the next file to the other disk.
It's configurable at what directory level the disk toggle works too.
On Reads, you do not have to know which disk the file is on because the userspace filesystem keeps indexes of all the files in memory and knows what disk it's on.
So over the last few months I've been playing with this and I like it allot.
I plan to use it as my backup system with HDUP (possbly backula) and media server.
I have two systems I'll be pimping in the near future, in the meantime here's unRAID!
PROS
Root filesystem in ram, very easy OS install. Drop in on a USB key, set bootable. reboot.
UI is browser based and very simple to set up.
Each disk is part of a protected array, Think of it as a protected JBOD. Like RAID4 without stripping.
User Space Filesystem allows SMB shares across multiple drives to be seen as one. I.E. it hides the fact that you have the same directories on multple drives and shows them as one.
Simple slackware 12.0 install. you can install packages from the net easily.
Boot filesystem filts on a 128MB USB key.
Handles single drive failure without loosing data.
Multiple drive failure does not result in loss of every single drive, only the failed drives vs raid5 and loosing the whole array.
Can use spindles of any size as long as the parity drive is the largest drive in the array. (good way to make use of those old drives)
Can be set to spin drives down when not used for a period of time.
If need be, you can remove a drive and re-use it, accessing the data on the drive from another linux machine (as long as reiserfs is compiled into the kernel).
CONS
All of the easy of single spindle drive management.
You can migrate a drive from a smaller to a larger with a few clicks or simply swapping out the drive.
All of the performance penalty of RAID5 writes without any of the read performance. (The price you pay for easy management).
Reads do not require the parity drive so you get read performance of a single drive.
Userspace Filesystem has write performance penalty.
No NFS yet. (coming soon)
No iSCSI yet (coming soon)
Still a bit immature.
Cannot use other linux raid options because thw whole md/raid subsystem has been replaced.
Limit of 16 drives + 1 cache drive.
Cannot concat or raid0 multiple drives for parity therefore it must ALWAYS be the largest drive in the array.
Posted by Me on May 1, 2008 11:00 PM