Fixstars High Capacity SSDs?

[MunkyTOS]

http://www.overclock3d.net/articles/storage/japanese_company_fixstars_make_10_and_13tb_ssds/1




Great googly moogly at the capacity

 

[Arbiter Odie]

That single drive almost has the same capacity as all the drives in my house combined.

Fixstar! I volunteer to benchmark and review the drive for you! And if you don't want it back, I'd be willing to hold onto it, if you made me.

I wonder what the pricetag is.

 

[Automata]

Wasn't there a thread where people said solid state drives won't replace mechanical drives? Yeah, never...

 

[Arbiter Odie]

I know, right? Here's seagate's latest and greatest
http://techreport.com/news/29597/helium-lets-seagate-store-10tb-on-its-latest-data-center-drive

10tb, in a 3.5" form factor. Whereas fixstar manages 13tb in a fat 2.5". Seriously.

Thideras, you should replace your file server drives with a couple of these. I bet the random i/o would improve a good bit, heh.

 

[Automata]

Thideras, you should replace your file server drives with a couple of these. I bet the random i/o would improve a good bit, heh.

I'd love to go full solid state for the storage array. The servers already are full solid state (8x 256gb Samsung Pros in each).

 

[MunkyTOS]

Im curious to see what price these will go for at launch given the storage per drive is ridiculous. Hell I'd be happy with 10-13tb mechanical drives

 

[deathman20]

Wow.... just need one of those 13TB drives and it would almost triple my storage capacity on my server. Grab another just for backup sake.

Now we'd be talking! Course its probably a few thousand dollars for something like that.

Nice to see that high capacity drives are coming up... which if truly the case yes it can beat/match mechanical drives on the storage side, just for one heck of a price for a while.

 

[wingman99]

I wonder when Samsung will come out with 13TB SSD.

 

[HankB]

First question. Who is Fixstars? Are they well known in the enterprise market? I've never heard of them in the consumer market.

Second... Half a TB through a SATA connection is OK. Is trying to squeeze 13TB of data through a SATA connection going to provide good performance? Relatively speaking, it seems like it is good for a lot of data as long as you don't need a lot of it at once.

And at what cost?

IMO if that kind of storage is needed, it would be smarter to RAID a bunch of one or two TB SSDs. Perhaps a bunch of NVME drives.

 

[EarthDog]

Squeeze data through sata? Is 580MB/s (reads) not enough? I'm not sure what would be squeezed with more capacity as throughput doesn't change.

Since this is enterprise, the point here is scaling and a small footprint. How many 1TB SSDs would I need for a petabyte? How many boxes would that take on my floor...more power use...etc.

 

[HankB]

Squeeze data through sata? Is 580MB/s (reads) not enough? I'm not sure what would be squeezed with more capacity as throughput doesn't change.

Here's what I was trying to get at. (If I can do the arithmetic correctly. )
A 1TB Samsung Pro SSD - at 550 MB/s that's 1000GB/0.55GB/s => ~30 minutes to read the entire contents of the drive.
A 13TB Fixstars - at 580 MB/s that's 13000GB/0.58GB/s => ~6 hours and 20 minutes.

Given sufficient I/O bandwidth, 13 TB of data on 13 x 1TB drive could still be accessed in ~30 minutes. I have no idea if such H/W exists.

I also realize that real world throughput is going to be less than the theoretical maximum and that depends on the I/O mix, file systems in use and probably other factors I'm not even aware of. But none of that can sidestep the bandwidth issue.

Since this is enterprise, the point here is scaling and a small footprint. How many 1TB SSDs would I need for a petabyte? How many boxes would that take on my floor...more power use...etc.

Enterprise is entirely outside my realm of experience. I would not be surprised to know that there are use cases where this drive would be better than any other existing drive. I guess I should phrased that as a question. Are there use cases for which this drive is not suitable due to the SATA interface speed?

 

[Automata]

The idea isn't to read the read/write the entire disk as quickly as possible, it is meant to store large amounts of data and retrieving it quickly when needed. You would see a considerable increase in performance going from spinning disks to these and the density is significantly better than existing solid state drives. In most situations, you don't need the best performance, you need the best density (disk count, physical size, and power wise).

 

[EarthDog]

+1 thiddy...

How many HDD would it take to reach that same speed? ~4 drives...then you have to worry 4 drives dieing vs one..to prevent that, put them in R10/0+1/etc. Exponentially more power use, weight, etc. If they needed more throughput for reading tons of data, RAID the ssds.

 

[Railgun]

Wasn't there a thread where people said solid state drives won't replace mechanical drives? Yeah, never...

Sure. Easy to think that when the price isn't known. I'm sure the $/GB is through the roof. That said, I've contacted them for a quote on the 13TB version.

 

[Automata]

Sure. Easy to think that when the price isn't known. I'm sure the $/GB is through the roof. That said, I've contacted them for a quote on the 13TB version.

A lot of research is being done into memory technologies for a reason, and we have a lot left to squeeze out of it. Mechanical drives will be extinct unless an amazing breakthrough increases capacity significantly.

 

[MunkyTOS]

What type of barriers are they trying to break through with mechanical drives?

 

[Railgun]

They came back and said it will be less than $1/GB...which probably means it will be something like $.80/GB, which means is still $10k. Not quite looking at the ~$.15/GB that HGST has for their 10TB spindle. Hardly the nail in the HDD coffin.

 

[EarthDog]

Hardly the nail in the HDD coffin.

Id say the nail is on the wood with the hammer ready to strike.... not that anyone said its a nail in the coffin, just that it is being phased out, and, it is.

Reminds me of the evolution in magnetic storage for enterprise... Reel to reel gave way to 3480 type tapes (210MB?) and now we are up to, well, the last I remember, LTO7 with capacities around 15TB, LOL! Yet, we, and many other companies, are going/gone tapeless. We have been a tapeless environment for YEARS. The speed that we can now back up directly to disk (spinners) is incredible compared to the throughput in tape... same with restore times as well.

 

[Railgun]

Fair enough.

Don't get me wrong, I'm all for it, but I don't think the crossover point is as soon as others think it is.

~5x price point is still a long way to go.

Let me give a steeper example. Seagates 8TB drive. $218 .02/GB

 

[EarthDog]

Its the speed though... at this time, not many need it... but you can bet, high I/O environments with lots of data such as CC companies, big banks, economic exchanges, etc, will be moving in that direction. When the price comes down, as it is, that will trickle down into smaller enterprise environments as well.

(also, I edited a bit more, sorry)