October 12, 201411 yr Ok, at $999 ($125/Tb) maybe not a good deal. But I could see this drive becoming affordable within the next year as Seagate and HGST (WD) try to win market share in the next generation of x-large drives. HGST 8TB Helium
October 12, 201411 yr Whoa that's a lot of money! Personally I'm just ready for the 4 TB hard drives to get very affordable. As I only have a 4 TB array currently.
October 12, 201411 yr Yes, I agree that $125/TB is "not a good deal" The 6TB versions of the helium filled units are down to $80/TB now ... at about half of that they'll be reasonably attractive. Actually, even at $50/TB they wouldn't be too bad. I'd pay a $10/TB premium for helium, but that's about my limit. Perhaps next year
October 13, 201411 yr You should be able to buy them at $799, or $100/tb in quantity 1. http://www.compsource.com/pn/0F23668/Hitachi-196/ But as you said, not quite a good deal.
October 13, 201411 yr Author I know it was a joke but I have an answer. Twice as long as a parity check with 4T parity.
October 13, 201411 yr I know it was a joke but I have an answer. Twice as long as a parity check with 4T parity. Only if the speed and areal density of the drives are the same (Even then, it'd be a bit less than that, due to the higher % of time the drive would be reading from the outermost cylinders) Maybe you could buy a few and let us know the actual results 8)
October 15, 201411 yr That's ignoring the 2-week badblocks test and preclear soon the 10TB version will be out if this increase of capacity keeps going up is there any way to increase parity checks speeds? or are we forever going to keep getting longer and longer checks.
October 16, 201411 yr That's ignoring the 2-week badblocks test and preclear soon the 10TB version will be out if this increase of capacity keeps going up is there any way to increase parity checks speeds? or are we forever going to keep getting longer and longer checks. A higher speed interface has already made its way onto motherboards and controllers, but these drives don't use it.
October 16, 201411 yr That's ignoring the 2-week badblocks test and preclear soon the 10TB version will be out if this increase of capacity keeps going up is there any way to increase parity checks speeds? or are we forever going to keep getting longer and longer checks. The short answer is yes, the checks are going to be longer as drives get larger => but perhaps not as much longer as you think. Remember that as areal density increases, so does the transfer rate from the drives. For example, the 4TB WD Reds have 1TB/platter density; the 6TB units are 1.2TB/platter ... so there's 20% more data transferred per rotation of the platter. As densities get even higher, the sustained data rates will also improve. But there's no getting around the simple fact that you have to read all the data to do a parity check.
October 16, 201411 yr That's ignoring the 2-week badblocks test and preclear soon the 10TB version will be out if this increase of capacity keeps going up is there any way to increase parity checks speeds? or are we forever going to keep getting longer and longer checks. A higher speed interface has already made its way onto motherboards and controllers, but these drives don't use it. I presume you're referring to the M.2 interface, which provides PCIe x2 (or x4 on the high-end version) interface speeds. This is certainly useful for SSDs, but would be meaningless for rotating platter units, as no drive comes even close to the SATA-3 interface speed ... and likely won't for the foreseeable future (even the fastest drives don't even get to half of the SATA-3 capabilities).
October 18, 201411 yr Yes, the 12Gb SAS units will be very nice as well. The M.2 units can achieve the same speeds ... and don't require a new SAS controller (although of course you have to have a motherboard with an M.2 slot [or 2 ].
October 18, 201411 yr I don't make arrays of 1 or 2 drives. :) ... None of us do. The M.2's are really nice for boot drives on a Windows machine, but clearly not useful for UnRAID. Clearly the new 12Gb/sec SAS controllers are better for that purpose. On the other hand, I suspect you also don't make arrays using SSDs -- and without doing that, the 12Gb/sec controllers are a waste, since no rotating platter drive in the foreseeable future is likely to even get beyond SATA-3 speeds in sustained transfer rates. An array of high-speed SSDs would indeed be VERY nice -- and as 1TB (and larger) SSDs get more affordable, it might even be a reasonable thing to do. I've thought about building an UnRAID Basic server with 3 small SSDs just for grins ... not for actual use, but as a testbed for new releases -- but I'll pass on a "real" array full of 1-2 TB units until the costs are significantly reduced.
October 18, 201411 yr I do make arrays with the SSD800MM and SSD800MH. I also can saturate the 12Gb/sec controllers with spinning disks.
October 18, 201411 yr I do make arrays with the SSD800MM and SSD800MH. NICE A bit pricey for my blood, but you've got to get some amazing performance with those!! ... I also can saturate the 12Gb/sec controllers with spinning disks. Even a 15,000 rpm SAS drive with 1TB platters (there aren't any drives with those specs) would only get a bit over 400MB/s sustained transfers speeds => still well under the SATA-3 rate. There's simply nothing available that will saturate these controllers with data coming off the platters. Clearly, all disks have a very tiny % of transfers that actually happen at interface speed -- the data that's in the drive's cache => but that's an irrelevant percentage of transfers, and has virtually no bearing on overall performance. Since a single drive can't come close, I assume you have some 12Gb compliant port multipliers that combine 4 or more drives onto a single SAS port ... is that how you're saturating this?
October 19, 201411 yr The fanout is usual more than 4. In the 70 bay system, it is 7 or 8 at different parts of the backplane. But large scale can go as high as 36 as described below. DataBolt Such Target (disk) bandwidth aggregation is the rationale behind what LSI calls DataBolt technology. Aggregation enables 12Gb/s SAS to be introduced into an existing 6Gb/s storage array in a way that provides an immediate doubling of overall system performance. In fact, with bandwidth aggregation, there is no need to use any 12Gb/s SAS disks to achieve 12Gb/s SAS performance at the system level.
October 19, 201411 yr The fanout is usual more than 4. ... Yes, I assumed as much. I only mentioned 4 because that's the minimum number of modern high-density disks that it would take to get close to the 12Gb/sec throughput (it would take 5 or 6 to actually saturate it). With even higher fanouts, the 12Gb/s is actually a bottleneck ... but, as you noted, it certainly would be saturated.
Archived
This topic is now archived and is closed to further replies.