SSD vs HDD. Harddrive vs PCI express lanes.

[FordGT90Concept]

Oh now i get it. So why they even invented that 12gb/s SAS controller anyway ???
Just to get those SAS SSD perform upto mark ? Still ssd i dont think have a life span better then spinning disk ? No ?

For 24+ hard drives.

 

[windwhirl]

Answering the OP question:

You could connect a hard drive straight through PCI Express. Is there a point? Not really. A PCIe 1.0 x1 lane has a transfer rate of 250 MB/s. Meanwhile, a 4 TB WD Black HDD (one of the fastest available on the market) can transfer at best 256 MB/s, as long as it is a sequential read/write and doesn't have to constantly jump from one track to another. So you'd be losing at most 6 MB/s of transfer rate.

Changing the controller wouldn't really matter. The speed limit is set by the very technology that's the foundation of hard drives: spinning platters and magnetic heads.

At best, what you can do is use a two or more hard drives in a RAID 0 or RAID 10 setup.

best shown example is Optical Drives they have peeked at around 50x ( much higher than that disks tend to spectacularly Explode).

I just had to look up "exploding discs" on Youtube... totally worth it lol

Oh now i get it. So why they even invented that 12gb/s SAS controller anyway ???
Just to get those SAS SSD perform upto mark ? Still ssd i dont think have a life span better then spinning disk ? No ?

SAS was never meant for consumer applications, in the first place. Second, I think you can connect multiple drives to one single SAS port (I've never seen this, since I don't really work with enterprise-class hardware, but I have heard of daisy-chaining drives with SAS).

 

[FordGT90Concept]

SAS was never meant for consumer applications, in the first place. Second, I think you can connect multiple drives to one single SAS port (I've never seen this, since I don't really work with enterprise-class hardware, but I have heard of daisy-chaining drives with SAS).

Using an expander, yes, but you're still limited by how much bandwidth the SAS controller has:

Serial Attached SCSI - Wikipedia

en.wikipedia.org

"SAS allows up to 65,535 devices through the use of expanders"

Important difference between SATA and SAS is that SATA is half-duplex (send or receive) and SAS is full (send and receive simultaneously).

 

[Aquinus]

Fun fact, centripetal force increases by the square of angular velocity, multiplied by both mass and radius. Hard drives can't spin infinitely fast for the same reason that traditional internal combustion engines can't. They also do the same thing to achieve higher angular velocities, by reducing mass and by shortening the radius of the spinning mass, which is why engines tuned for high RPMs have a very short stroke lengths and why high RPM hard drives have smaller platters. Less mass and a smaller radius reduces the effects of the centripetal force, however since it's the square of angular velocity, it imposes an upper bound on how fast these things can rotate at while not becoming a paperweight.

 

[aQi]

Fun fact, centripetal force increases by the square of angular velocity, multiplied by both mass and radius. Hard drives can't spin infinitely fast for the same reason that traditional internal combustion engines can't. They also do the same thing to achieve higher angular velocities, by reducing mass and by shortening the radius of the spinning mass, which is why engines tuned for high RPMs have a very short stroke lengths and why high RPM hard drives have smaller platters. Less mass and a smaller radius reduces the effects of the centripetal force, however since it's the square of angular velocity, it imposes an upper bound on how fast these things can rotate at while not becoming a paperweight.

Western digital velocity raptors is a clear example.

 

[Aquinus]

Western digital velocity raptors is a clear example.

Exactly. They're basically the same size as a laptop hard drive for a reason. Less mass and shorter radius. As you can imagine though, that limits capacity.

 

[aQi]

Exactly. They're basically the same size as a laptop hard drive for a reason. Less mass and shorter radius. As you can imagine though, that limits capacity.

Though high rpm with less mass produces heat so they ended up with ice cage (heatsink) with them..

 

[Aquinus]

Though high rpm with less mass produces heat so they ended up with ice cage (heatsink) with them..

I'm not sure how necessary that is, but it's likely due to heat caused by friction in the bearing, heat emitted by the motor, and possibly by the air turbulence inside the drive itself. As you can imagine, everything gets harder to manage as a rotating medium increases in speed. That's why for high performance applications it's a far better idea to get an SSD because no moving parts. However, HDDs still are the winner when it comes to capacity (for both size and cost.) Although that gap is slowly closing.

 

[R-T-B]

Density is an HDDs only remaining advantage. That, and in some VERY SPECIFIC work cases, endurance.

 

[newtekie1]

Density is an HDDs only remaining advantage.

Not even really density, as I believe flash is now capable of storing more data in the same amount of physical space.

The only advantage HDDs have now is $/GB.

 

[FordGT90Concept]

Physical space isn't the point so much as magnetic fields are physically smaller than transistor gates.

3D XPoint is the only memory that comes close to the endurance of hard drives. It's still far behind.

 

[Bill_Bright]

Leave the Sata barrier aside. Im taking about a controller that would use spinning disk instead of nand flash storage and provide optimal read/writes through pci express lanes

You can't leave any barrier/bottleneck behind.

And while scanning through this thread, I didn't see where the actual writing of the data to disk was addressed. That takes time too because the R/W has to physically (via magnetic impulses) reorient the magnetic particles on the disk. As they get smaller and smaller (with greater densities) reorienting takes less time, but it will never be as fast as flipping a 0 to a 1 with SS storage devices. This is why the slowest SSD can run circles around the fastest HD, even SSHDs.

It does not matter how much data and how fast that data comes through the interface - actual reading and writing the data will always be a big (tiny?) bottleneck.

And of course, it takes time for the read/write head to swing back an forth then properly center itself over the track - even after the disk is defragged.

as long as it is a sequential read/write and doesn't have to constantly jump from one track to another.

But constantly jumping from track to track is inevitable and unavoidable. Even if the drive 100% defragged (not likely) every file needed will NOT be in sequential order. And most files are tiny so the R/W head will have to jump back and forth to get to the next file location.

Physical space isn't the point so much as magnetic fields are physically smaller than transistor gates.

Yes, but that creates a problem in itself. It still takes energy to flip and flop those gates. But you can't have those magnetic fields interfering with the state of any adjacent gate. So field has to be tiny enough to affect only the intended gate but at the same time, strong enough to physically flip (or flop) it. That takes some extremely precise positioning - and that also takes time.

I agree with

Spinning Hard Drives will be here for years yet.

But they will be here ONLY because they still offer more storage for the $$$. As soon as SS storage becomes cheaper than HD storage, HDs will rapidly go the way of the 8-Track. And note - especially when it comes to data centers - it is not just about the cost of the drive. HDs consume more energy, generate more heat, weigh more, take up more space, make noise, and of course, don't perform as well.