Back in the beginning of 2008, I purchased my first SuperMicro 846 24 bay chassis with dual 900W power supplies. I loaded it up with 24 1 TB Seagate Enterprise drives, all controlled by an Areca 1170 RAID controller running RAID 6.
Here's the chassis:
And the internal wiring:
In August 2009, I upgraded the mobo to a SuperMicro x7dwn+ running Windows 7 Ultimate.
At this time I also added a Areca 1680 controller and a SuperMicro 846E 24 bay chassis, loaded up with 2 TB Hitachi consumer drives. This was also setup as a single RAID 6 array.
Over the last 4 years, I lost maybe 3 of the Seagate Enterprise drives (ST31000340NS), in the 24TB raid, but I always kept spares on hand and rebuilt the raid no problem.
The 48TB raid however, has given me several issues. Sometime during a reboot (generally only from extended power outages), the controller will loose the chassis, and I have to fiddle with the cables and do multiple reboots to get it back. I have also lost about 10 drives now (fortunately all replaced under warranty), but that is about to expire and the failure rates have been steadily increasing to the point where I'm loosing a drive now on every other rebuilt.
So, I'm thinking it's time to move from these colossal enclosures (which also consume huge amounts of power with 4 900W power supplies running 24/7).
I'd like to move to independent 1 or 2 U chassis, each containing 6 drives in a RAID 5 configuration and then network raid them all together across my network and have each "node" on the network go to sleep when not in use and use WakeOnLan. I'm using MyMovies btw, for my rather large collection of Movies and TV Shows, btw.
So any suggestions on what's available, preferably free that I can hack together myself?
Just to clarify, what I envision is that each RAID5 network node will connect over the network to form a single virtual volume that I can then map to all my HTPCs. I figured I'd need at least 6 nodes to start in a "network RAID 6" configuration. This way I can loose 2 entire nodes at the same time and the virtual volume would stay up. To loose a node would require me to lose 2 of the 6 drives in a particular node, or for it to have some sort of other problem that would take it offline.
Going network RAID for the sole purpose of aggregating your storage is a rather terrible choice.
Since the disk aggregation is done at the block level, any network issue would wreck havoc.
Network RAID is not for simpleton home networks as you will pay a very high price for latency unless it fully runs async. Great in theory. In practice, its requirements for trouble free operations are stringent.
You can certainly go with hardware RAID on each node, but then I would recommend that you export each volume as a network share and then pool all those shares using FlexRAID's RAID over File System pooling, which is the only solution capable of pooling network shares.
For media storage, I would even question going hardware RAID, but then I could be biased.
Start thinking of having only a single disk spinning when watching a movie on that massive array of yours while all other disks are in a sleep state. Then think of how a disk failure past the tolerance level will affect only the failed disks and not the whole array. Think of being able of mixing disks of any sizes, makes, or models. Online RAID Expansion and Contraction. SMART monitoring with email and SMS notification.
You might even go back and have them all in one energy efficient system + attached storage for the overspill.
Do yourself a favor and download Transparent RAID
, save a few trees, and simplify your life.
Appreciate the advice!
I use network raid on a large scale at work (various Lefthand/HP/Dell systems) and was curious if a consumer level solution existed.
Transparent RAID looks very interesting! I'll have to look into that further for sure. Fortunately my Areca controllers can be configured as JBOD (meaning that each drive is just in pass-through mode).
I currently have the RAIDs set to spin down after 60 minutes of not being accessed. I also have them set to spin up in a staggered mode (0.2 second delay between spindles). Furthermore, each RAID is set to check the volume every 2 weeks after 30 minutes if idle time. It basically carries out the following tasks:
1. Scrub bad block if bad block is found, assume parity data is good
2. Re-compute parity if parity error is found, assume data is good
But despite all of that, I have been very close to loosing an entire raid several times and years worth of ripping and organizing effort. I really like the advantages of tRAID over my rather rigid RAID6 solution which I'm always worried about having a catastrophic failure. Being able to pull out a drive and read it in another machine is great!
Have you considered a Drobo appliance? I've owned two...a Drobo V2 and a Drobo 5N. I love that they are "hands off" management.