so I was looking at upgrading my unraid server. I was thinking of a board that when unraid gets updated to support more drives then get it now.

 

I have known about this board for some time. it has 5 x16 slots abd 1 x4 slot so with that supermicro 8 port raid card 8X5=40+8 onboard 48 drives. 1 slot is for a video card

 

what you think?

I'm not sure its worth investing that much in a MB. By the time UNRAID supports that many drives, MBs with have changed dramatically.

First of all, there's no such thing as future proof, especially in the world of computers. Ya pays your money and ya takes your chances.

 

I think that unless I had an urgent need to upgrade now, I'd wait until I could get a board with UEFI rather than BIOS. We should start seeing them next year. UEFI is what will allow boot devices greater than 2TB, and recognition of GPT rather than MBR devices natively, rather than with OS interference.

 

I'd also not bother with that board, or anything on that chipset. If I were going Intel and had to buy something right now, I'd go for the H55 chipset, because it supports the onchip video capabilities of the Core i3, where P55 doesn't. H series boards are also cheaper by a fair margin.

 

Finally, why would you even think about that many drives with a single parity drive. I doubt unRAID will ever support that many drives in a single server. I'd rather build multiple servers than to contemplate that. Just imagine the parity calculation times. And the monster PSU you'd need. And the wiring headache. And what kind of case you'd need. It goes on and on.

Although I agree with pretty much everything kenoka stated, the Supermicro offers several motherboards that will support many drives and have a proven track record (and lower cost). The X8SIL series will support 30 drives and the X7SBE will support about 46 with the same performance. I don't know what the future will bring, but I agree that we will probably have to buy new hardware to support 3T. drives. I'm hoping it will be just a controller card, but only time will tell.

 

The only way a array that size becomes feasible is with a second parity drive.

The reality is that no one who understands the math behind drive failures should be comfortable beyond 20 drives in any drive array, regardless of the parity method.  I myself cringe at getting much beyond 12 drives.  Why?  Because the likelihood of having multiple drive failures goes up geometrically as you continue to add drives inside a single array.  The correct solution, for the time being, is multiple arrays with independent parity.  The best solution I've seen is the proposed q-parity, but at the moment that is still Borg tech.  So instead of planning for a 48 drive system, I would look more into the reliability of the components in the system.

The reality is that no one who understands the math behind drive failures should be comfortable beyond 20 drives in any drive array, regardless of the parity method.

 

The likelihood of having multiple drive failures goes up linearly as you continue to add drives inside a single array. The events (drive failures) are independent.

 

 

The reality is that no one who understands the math behind drive failures should be comfortable beyond 20 drives in any drive array, regardless of the parity method.

 

The likelihood of having multiple drive failures goes up linearly as you continue to add drives inside a single array. The events (drive failures) are independent.

 

 

What we really care about is the likelihood of concurrent multiple drive failures, not individual failures.  The concurrent failure rate depends on how long it takes to identify a failure, replace, and re-construct any given failed drive.  If you estimate a given drive will fail in a 5 year period, and it takes you 5 days to identify/replace the failed drive and restore parity protection, you just need to figure the odds of another drive failing in that same 5 day window.  The more drives in a given array, the more chances of a failure in a given year, and the more chances of a concurrent failure in the same window of time where the array is working in a degraded mode.  I don't think the odds increase linearly with increased number of drives, but I could be wrong.