Seagate doubles down on HDD speeds with Multi Actuator tech

https://newatlas.com/seagate-multi-...ail&utm_term=0_65b67362bd-58c9a63dbb-92435577

John.

 

With Seagate leading this new tech I'm positive I won't touch it. Without doubt it will bring about a laggy (or at the very least; a highly variable responsiveness) and unstable aspect to any system it is used in. Unless that system is heavily tweaked (i.e. 'future data centers...') to use it specifically, this is an easy pass for any other usage style.

What Seagate is proposing here doesn't change the base speeds of the drive. This can't/doesn't magically change the read and write speeds of single files, even if it theoretically should speed up queued requests that the O/S can use in any order they are received.

The way to truly get faster performance from a HDD and/or an SSD for that matter is to design each in such a way so that they can read and write at the same time (i.e. full duplex). For HDD's, that means at least a read and a write head for each platter. The next bump up in performance will be to have the platters behave as if in a RAID0 array - with large enough files written (and read) to as many platters as possible, consecutively.

But given the power requirements, none of the above will come to pass anytime soon in the HDD space.

I'll certainly read the reviews if/when these come out.

I'm not saving any pennies to buy these with anytime soon though...

 

This isn't the first time they've tried this: https://www.tomshardware.com/news/seagate-hdd-harddrive,8279.html . It didn't make a lot of sense in 2009 and makes rather less now. The performance benefits were outweighed by the increased complexity.

Doubling the IOPS of a hard disk then was interesting. SSD's were much lower capacity than hard disks and tremendously more expensive. Now, it's not. We have higher capacity available in solid state -- 8 TB in 2.5" 7mm form factor and 100 TB 3.5" -- with 2-3 orders of magnitude more IOPS; a factor of 2 won't make much of a dent in that. Yes, the bare drives are at least 5x the price of the spinning rust variety, but the drives themselves are only a small part of enterprise storage cost. Big capacities mean higher density (less rack space) and fewer controllers needed.

HDD sizes have stalled out over the past 5 years. 2.5"/7mm is stuck at 2TB with no signs of moving forward. 15mm drives, to be sure, have increased -- from 4TB all the way to 5TB. SSDs, particularly very large ones, are still expensive, but that dam will break one of these days.

 

This isn't new tech. They also did it in the mid-1990s (~1995 or so) with the Seagate ST12450W, Barracuda 2 2HP which was a dual actuator ~1.7GB drive. Only available in Wide-SCSI. Cost about $1800 or so.

It had awesome performance for the time, nearly double that of the drives that were offered at the time, and in terms of IOPS never really has been succeeded since (although some of the 15k Seagate Cheetah's came pretty close!). But it proved to be too expensive relative to the alternatives such as RAID-0.

I don't see why Seagate is bothering this time around. Anyone who needs higher IOPS has already moved to SSDs, making this a rather expensive gimmick.

 

Seagate didn't actually build anything in 2009. Last dual actuator drive they built was in the mid 1990s. A 1.6" high, 11 platter monster. Which had IOPS performance that 20+ years later still hasn't really been superceded by a HDD.

 

Could be; I thought I remembered something newer, and that's what I found. I wouldn't expect newer HDDs to have markedly better IO/sec; seek times haven't changed, and while higher rotational speeds help, that's only part of the equation. Those puppies were 7200 RPM; there aren't a lot of faster drives these days because anyone who wants better latency just goes with SSD and done with.

 

I’m sure the reliability will be less than stellar...pass

 

This is kinda interesting. So long as it doesn't have too big of an impact on price or noise, and it's reliable, I might be interested. Realistically I'm going to have hard drives around for mass data storage for awhile, and being able to access that faster (even if the IOPS still pales in comparison to an SSD) would be nice.

Of course too much of a price increase and it's not worth it over an SSD... but we're at about 2.7 cents/GB for HDDs and 27 cents/GB for SSDs right now, using Toshiba's 7200 RPMs and mass-market SSDs as examples. There's a place that it can slot in.