WOW, 15k rpm HDD from Fujitsu (News)

I spotted this on TCM and wondered if we can fry an egg over it

 

Great, now your laptop can vibrate more than my washing machine...

Bad jokes aside, SAS drives are meant for servers; they cannot be used in laptops due to excessive power consumption, heat, and vibration issues.

 

Yes, I know that. Just wonder how hot they are

 

And SAS doesn't have standard connections as notebooks have.

 

And Fujitsu is now Western Digital....

 

That I did not know.

 

Yes, Fujitsu is negotiating to sell the HDD production plant for WD.

 

SAS and SATA are interoperable to an extent. You can hook SATA up to a SAS controller. You can't hook a SAS drive up to a SATA controller. SAS connectors are standardized, but not compatible with notebooks.

I think it's just less than optimal wording on your part.

 

there wouldn't be enough power to power the drive anyways, even if you could get it connected

 

Its great and all but these are not working in your notebooks, so dont get your hopes up.

K-TRON

 

That's what i meant, can't hook a SAS-drive to a SATA-controller.

And what do you mean with that last sentance?

 

You said:

That's inaccurate SAS connectors are standardized otherwise it wouldn't be a standard They are slightly different from SATA. Based on where you're located it's likely that you aren't a native English speaker - which is why I was trying to be gentle with my wording.

I don't know that that comes across with my typing.

 

Hehe, now i see what you mean!

And yeah, you're correct. It was not an optimal word that i chose.
I'm not a native English speaker at all, i'm just a swede

 

I am sure these will carry quite a price, but I'd love to stick them in a server.

 

What I don't understand is why do they make it 2.5'' instead of 3.5''.

 

Higher disk density to make it even faster.

 

Several reasons all related to the fact that a 2.5" platter is smaller than a 3.5" platter.

1) Average and maximum seek times are shorter as the worst case scenario has the heads seeking a maximum of 1" less.

2) More drives into the same physical space. Some of my 2u servers have capacity for 16 2.5" drives. I would only be able to fit about 10 3.5" drives.

3) The platter weighs less so it requires less power to spin. Comparing Seagate's previous generation 15K 2.5" versus latest generation 3.5" it's nearly a 2:1 power savings in favor of the 2.5" drive. I can't find specs for the newer generation 15K 2.5" drive. Less power == less heat to dissipate.

That's a good start for why.

 

More than that. See my post below.

 

Just keeping it simple.

 

Any drive at any RPM will see a speed bump with higher areal density; the bits move under the head faster. This is why you will see larger hard drives in the same form factor outperforming lower density drives with all other factors being equal.

On long sustained transfers, a 3.5" disk at the same speed could see higher throughput as there are less seeks to move to the next track.

Simplistic answers are fine as long as all you want to do is scratch the surface.

 

Since these drives are made for speed I believe my answer is very fitting, thanks. Also sustained transfers has no benefit when you are working with files such as databases.

 

The areal density differential is a non-factor on database environments. OTOH, faster track to track and full stroke seeks are a very significant factor.

 

I said sequential speeds did not come into play when you are working with smaller files which you need to access quickly. Indeed disc density would not really factor in so there would be no point to use say a 1TB drive where the disc density is higher. That is why you use these types of drives.

 

keep pushing the HDD when the SSD technology will own .. soon

 

I think the most noteworthy thing about these drives is that they're SAS 2, rather than plain old SAS. They're the first products with SAS 2 that I've seen. That means up to 750 MB/s, rather than the 375 MB/s from SAS, 300 MB/s from SATA 3 GB/s, or 640 MB/s from Ultra SCSI-640. So if you put enough of them together in a RAID array you'll be able to significantly faster times, especially relative to SAS/SATA.

So if you max out this interface with these drives you'll have hard drives that are almost as fast as PC-100 RAM in throughput. Which for a hard drive is mighty fast.

And the interface could be relevant for SSD's as well - if they release SAS 2 SSD's, they could have much higher theoretical speed than current SSD's when RAID'ed.

 

Of course harddrive technology is going to continue. A mechanical drive is more reliable than a ssd. Enterprise drives have proven themselves, and the are durability is why they are used in Servers. SSD's have not been out very long, and surely they will not last as long as a mechanical drive.
Sure a SSD is faster, and wont break if you drop it, but mechanical drives have been perfected, and they will run for years without any problems. (unless you cause problems)

K-TRON

 

lol, no way

 

I hate to use that word twaddle again K-TRON, but I think it's required for these two comments.

The NAND flash memory cells found in SSDs can last for years beyond the three to five year life expectancy of a magnetic hard drive. Because hard drives include numerous moving parts, they are vulnerable to wear and tear over time.

An SSD can still break if you drop it, but as a whole the lack of moving parts make them much less prone to damage. If left un-bothered, a solid-state drive can last up to 60 years longer than a hard drive in a similar desktop environment.

Once these initial kinks such as inferior random write and sequential write times are ironed out and the prices come down they will devour the HD market.

 

yeah, you can relay the information you read on the net, but it hasnt been done. No ssd has been around for 5-7 years, so you cant just say its going to, cause guess what you dont know that for a fact.
Mechanical drives are still the way to go

K-TRON

 

flash has been around for more than 5-7 years, and military uses ssd's since over 10 years.

flash based industry pc's are around since win95 (and longer ). and i know some that still do their work. (and yes, i've worked on them and had to repair the win95 installation from time to time..)

so you're definitely wront k-tron.

no one really cares about reliability on the performant system drive anyways. do a backup and use a system drive that works well for 2-5 years and then you don't have to ****** care at all. because after those years you'll buy a new system anyways with a new much faster ssd for much less. so what?

you just lost all your other points so you point out the "they're new, don't trust them".

 

Okay mister smartypants, than why arent they used in servers?

K-TRON

 

lol simply, they're not used in servers because they don't offer enough space

 

I am very close with many enterprise computers supply companies and the orders for SSD are increasing. A high speed SLC SSD (16-32GB) for the OS is what customers want.

 

because "never touch a running system" and servers most of the time don't need that performance.

for the servers where it mainly matters, solutions like ssd or ramdisks where used before.

ssds enhance the server performance. but a lot of companies don't bother much about that (if the cost is no direct benefit like making more money).

here at work several systems are very slow and would benefit from an ssd storage. but if they're fast enough to be useful, why spend money on them?

 

oh, and a lot of highperformance systems are designed around harddrives. see sql server. it eats up as much ram as you want to give it and runs in the ram then. the harddrive then will be only sort of a backup.

tons of gb of ram are not that expensive as tons of gb of ssd, and they're even faster. but the price/benefit ratio is drifting. so instead of a highend harddrive like those 15k rpm disks (that often don't offer _that_ much storage as well) one can spend the money for an ssd, too.

 

SSDs are just starting to get a foothold in servers, I know of a few companies that sell them for OS drives or database work and then have additional drives for storage. Though right now they aren't as cost effective or a proven technology yet, so most would rather go with something like this drive.

 

did you guys know that fiber optic cables have been around about 30 years before those became available for the public ? ... I bet there have been SSDs when I was using DOS 3.0

but anyways, back on topic.

 

There have been... SSDs has been around for at least a decade

 

Right.

That's exactly where we're headed -- the OS will go on SSD. The "big storage", i.e. the Apps and DB go on massive redundant storage arrays with even more disk to back those up. Tape is literally for long-term archival storage for legal purposes.

Remove one of the most likely items to fail inside my chassis from the equation? Absolutely. That they run cooler is "just" a side benefit.

 

You still have to flush RAM to disk on a fairly regular basis. If you don't, one power blip and you've lost tons of data. Not good on an enterprise level DB.

I don't consider sql server to be a part of a high performance system I think I work in a different world of high performance servers than you do.

You're still looking at orders of magnitude differential in throughput.

I see even small systems' storage going to a multi-tiered solution

(1) Cache
(2) DRAM
(3) Solid State direct via a bus
(4) "Traditional Hard Drive"

Traditional hard drive would be for the big storage that isn't speed sensitive.

 

well it all depends on the needs. but generally, the only thing where it matters for a company to have quite high performance is in databases (i just took sql server as an example). there may be special needs where special performance requirements are needed, but i'm thinking of the "ordinary business".

and yes i think the multi-tiered solution will get much more used in the future. with sas2, most of the stuff could be done with very little ram, as caching isn't that needed anymore (and you get power-savety).

if it's mission-critical you shouldn't ever have anything in ram anyways as you correctly stated. the sql ram caching is mostly for databases with mostly read-needs. the typical example of the webshops come to mind.. most want to quickly find the right products. if the actual ordering then takes a little more time to commit, that doesn't bother anyone.

the other thing is, more and more the disk shows up as a bottleneck. before the multicore superservers, most servers where stuff like p3 1ghz dual cpu, and there, a slow disk wasn't always the problem if you now have some quadcore 2-3ghz cpus, you'll have >16x the performance of those p3. but in no way you can scale up the disks 16x with ease. except with ssd's. espencially in the latency case, one can simply not get more out of disks..