BackBlaze Storage Pod 3.0 [March 2014 Update: Now it's at 4.0]

[neilt0]

They've just published the specs: http://blog.backblaze.com/2013/02/20/180tb-of-good-vibrations-storage-pod-3-0/

 

And there is plenty of useful info for us builders there as well.

 

[aiden]

I'm a Crashplan user myself, but I love the fact that these guys continue to publish their designs for free. Their experience and deep pockets helps us design our systems. Besides that, it's just freakin' cool. 

[caseyparsons]

Besides that, it's just freakin' cool. 

+1!! That is SO cool. Wish I had the time and money to build a few of those kickass boxes.

[bnevets27]

Good info!

 

Seagate – ST3000DM001?9YN166 (slightly higher failure rate)

 

Unfortunately for me I just added one of these to my array as parity. Oh well that's why we have redundancy.

[JonathanM]

Good info!

 

Seagate – ST3000DM001?9YN166 (slightly higher failure rate)

 

Unfortunately for me I just added one of these to my array as parity. Oh well that's why we have redundancy.

At least it's parity and not data. Parity is the most important speed wise (all writes to the protected array are effected) and the least important failure wise (no data is directly in jeopardy if it fails). Any two drive simultaneous failure will lose data, but if one of those drives is the parity, you only lost 1 drive worth of data. If two data drives fail, you lost them both, and parity is worthless as a recovery.

[bnevets27]

Good info!

 

Seagate – ST3000DM001?9YN166 (slightly higher failure rate)

 

Unfortunately for me I just added one of these to my array as parity. Oh well that's why we have redundancy.

At least it's parity and not data. Parity is the most important speed wise (all writes to the protected array are effected) and the least important failure wise (no data is directly in jeopardy if it fails). Any two drive simultaneous failure will lose data, but if one of those drives is the parity, you only lost 1 drive worth of data. If two data drives fail, you lost them both, and parity is worthless as a recovery.

 

Very true, I kind of knew that going into it with that drive (had read that it might be a little more prone to failure) but it was the cheapest at the time and 7200 and I knew I was going to be using it for parity. So not all that bad. Curious which drives they prefer. Probably which ever is the cheapest at the time due to the amount they purchase.

[neilt0]

Curious which drives they prefer. Probably which ever is the cheapest at the time due to the amount they purchase.

No, they are very much in to reliability as well as cost-effectiveness. Earlier, they posted stats on drive failures and the Hitachi drives had a much lower failure rate than any other.

[vca]

I found this note on vibration rather interesting:

 

The key advantage of the drive bay assemblies is to reduce vibration. These assemblies not only keep the drives still, they also keep them firmly seated in the backplanes. Over the past several months we have tested different models of drives in the new drive bay assemblies and we have seen a dramatic improvement in overall system performance along with lower drive failure rates.

 

I can see how vibration could cause connectors to work loose causing an apparent drive failure, but they appear to be implying that the vibrations might be causing a reduction in drive performance as well.

 

Regards,

 

Stephen

[neilt0]

I found this note on vibration rather interesting:

 

I can see how vibration could cause connectors to work loose causing an apparent drive failure, but they appear to be implying that the vibrations might be causing a reduction in drive performance as well.

 

Regards,

 

Stephen

 

[korith]

I found this note on vibration rather interesting:

 

I can see how vibration could cause connectors to work loose causing an apparent drive failure, but they appear to be implying that the vibrations might be causing a reduction in drive performance as well.

 

Regards,

 

Stephen

 

 

Alright now that was very interesting. I wonder what effect a sustained vibration would cause.

[bnevets27]

I found this note on vibration rather interesting:

 

I can see how vibration could cause connectors to work loose causing an apparent drive failure, but they appear to be implying that the vibrations might be causing a reduction in drive performance as well.

 

Regards,

 

Stephen

 

Wow....that's mind blowing.

 

 

[shat]

IFL

[PeterB]

... they appear to be implying that the vibrations might be causing a reduction in drive performance as well.

 

What the Sun Video is showing is that it's the latency, or seek time, which suffers.  I guess that this is understandable - the head is moving across the disk surface, trying to position accurately onto one track.  If the whole device is shaking, then the head may well over or undershoot, requiring a read to establish actual track position, followed by a step to the correct track.

[c3]

Vibration has a multitude of negative impact, reliability, performance, power, even noise (a positive feedback loop). In a chassis like backblaze 1.0, traditional RAID would fail out drives for response timeout. The vibration effects heads, forcing them off track, requiring reseek, since the entire stripe is delayed, the drive is dropped. All the reseeking drives up the power and adds vibrations.

 

This area has a lengthy patent list...

 

[vca]

Vibration has a multitude of negative impact, reliability, performance, power, even noise (a positive feedback loop). In a chassis like backblaze 1.0, traditional RAID would fail out drives for response timeout. The vibration effects heads, forcing them off track, requiring reseek, since the entire stripe is delayed, the drive is dropped. All the reseeking drives up the power and adds vibrations.

 

This area has a lengthy patent list...

 

So I guess we don't want to locate a rack of disks in a tap dance studio :-)

 

Stephen

 

[aiden]

 

This is why I stopped yelling at computers. Humans are a much softer target anyway. 

[aiden]

I was just curious if anyone has successfully tested the port multipliers in this design with unRAID?

[neilt0]

BackBlaze Storage Pod 4.0: http://blog.backblaze.com/2014/03/19/backblaze-storage-pod-4/

[MyKroFt]

Will unraid ever support 45 drives?

 

Myk

 

[jumperalex]

I want two parity protected 24-drive arrays ... arranged into a mirrored array >:-}

[JonathanM]

Will unraid ever support 45 drives?

 

Myk

Tom has said it would require an almost top down rewrite, because of the 4 character drive designations required. (/dev/sdaa etc.) So... probably not. It seems like the current drive limit matches the physical characteristics of relatively cheap server hardware, moving beyond 24 drives is pretty expensive. Plus, with drive sizes continuing to increase regularly, I think the main concern is adding protection with dual parity. (insert large drive size here) is a crapload of data to lose. Increasing the spindle count increases the risk of dual failure more than I'm comfortable with.

[mrow]

.... I think the main concern is adding protection with dual parity. (insert large drive size here) is a crapload of data to lose. Increasing the spindle count increases the risk of dual failure more than I'm comfortable with.

 

I totally agree. I decided a while ago I won't add more than 10 disks while we still only have a single parity disk. 4TB, and now 5TBs, is a shitload of data to lose if two disks fail. If my storage needs ever outpace how much can be stored on a 10 disk array of the largest size disks I'll just use my currently unused second license and setup a second array.

[apgood]

Same here I reduced my array to 8 data disks because I did have a two disks go bad after a short in a shredder tripped the mains fuse.  Luckily I had backups of all photos and documents (and managed to copy most things off of one of the drives and rebuild sort of the other),  so only lost a few movies and corrupted a few more.

 

Double parity is definitely high on my wishlist for feature enhancements for unraid.

 

Sent from my Nexus 4 using Tapatalk

 

 

[neilt0]

IIRC, they run multiple RAID6 arrays on each 45-drive box, it's not one array.

 

Sent from Moto G using Tapatalk

 

 

[jumperalex]

Right.  And I've said before, and others I think too, that the fastest way to sort of get what we want is to add the ability to create two arrays in a single instance.  For now, it would just be possible to use Xen to create another instance of Unraid but that means using pass thru of SATA and as we've been told that is not ideal.

 

But dual array in a single box would be the fastest way to >24 drives and to afford more parity protection without the efforts to create p+q or some other dual parity method.

 

Sent from my XT1060 using Tapatalk