So IBM wants to kill their own creation: the hard disk drive

I came across this interesting article today: IBM wants to kill the hard drive it invented. Called racetrack memory, IBM is leading an effort to more affordable and higher performing non-movable storage. While the article was an exciting one and does give us insight into what IBM is planning, I wanted to spend some time explaining why I am surprised that we are still using the Hard Disk Drive (HDD) and haven’t officially moved completely away from this old technology and on to the better Solid State Drives (SSD).
As it has been pointed out in a recent post by the Storage Swiss, the SSD is a very predictable technology. For most of us in the computer industry, we are not strangers to the HDD failure. As soon as that clunking noise is heard, we scramble to salvage whatever remains of our data and in many cases replicate it to another drive. The thing with the HDD is that it is entirely mechanical with always moving components. Calculating its failures is a near impossible task. Albeit through trial and error, the SSD has evolved where the controller’s firmware can now predict Flash wear to even life expectancy. In many cases, it can overcome some of these scenarios.
Energy Consumption
No two SSDs are alike. Some demand more power than others which is why specific vendors (such as Micron or HGST) will market these different drives to fit specific environments. That aside, your traditional SSD will most likely consume significantly less power than a spun up HDD. The reality is, most storage systems to even system administrators rarely if ever spin down idle drives. A spinning drive will continue to consume power despite the drive(s) not being used.
Reduced Noise
Another advantage to not having movable parts is that you do not have to hear [and feel] the platter rotate on the storage device. A rotating drive is a vibrating drive. Typically drive enclosures are built to limit this vibration but just enough ekes out to the point where you can feel it when placing your hand on the front of the chassis holding those drives. A rotating drive also makes a lot of noise, when it is spinning up, spinning down, to even seeking across multiple platters to retrieve randomly dispersed data. I do not know how many of you have been in a data center but to those who have, I am sure that you will not disagree with me in saying that it is extremely noisy. Air conditioners blowing. Chassis alarms beeping in either rhythmic or a continuous non-stop tone. Servers and storage enclosures are running 24/7 and their fans are continuously cooling them down. Then we get to the Hard Disk Drives which as stated above is a contributor to this symphony.
Watch this video on the effects of shouting on a storage array filled with your traditional and age old spinning magnetic disk drives. Go on, I will wait here in the meanwhile.

Scary, huh?
Yes. The standard SDD is more expensive per Gigabyte than your traditional HDD. Although, it is basic economics: an increase in demand will drive an increase in production which in turn will decrease costs. Besides, determining how the costs will affect you is no easy task. You need to compute the cost of productivity between both types of media. Does it make a difference if the end user reads / writes data quicker? Also, back to the point of energy consumption. The SSD consumes less power than your average HDD, which would mean a slightly lower electric bill. They also generate less heat, reducing cooling costs, which could equate to larger savings on that same electric bill.
My Two Bytes
Here is the thing. Failures do happen with SSDs. I am no stranger to these kinds of failures. For instance, many years back I had purchased an OCZ Vertex II (using an LSI SandForce SATA controller) and within 45 minutes of use (right out of the box) the controller failed. That is, the PC it was plugged into was no longer able to detect it. It was dead.
Failures also occur with your traditional HDD. The difference is, the NAND and SSD controller technology has evolved significantly over the past few years. Aside from capacity and interface type, the HDD has not. The HDD you use today is really no different from the HDD used a few decades ago.
Either way, you need to be smart and prepared for when these failures do occur. Backup your data and do it frequently. Replicate the drive or group a bunch of them into some sort of redundant array.
Still not convinced on why you should migrate to SSDs? To quote a recent advertisement Fusion-io posted on Twitter for their newly introduced Atomic Series and why you should invest in these technologies: “Because latency is the new dial-up.
Comments are always welcome.

2 thoughts on “So IBM wants to kill their own creation: the hard disk drive

  • Your arguments are not new. These same promises of a SSD takeover have been made for 20 years. Hard Disks continue to be more cost effective. It is not guaranteed that flash will ever catch up. Flash is encountering it’s own technical challenges slowing growth.

    • Brian,

      You are absolutely correct. The arguments are old ones, although there is a large difference between the Solid State technologies of 20 years ago and today. What Texas Memory Systems was pushing out back in the 80’s and 90’s was predominately DRAM and at very low capacities. It was a lot more of a niche market technology. Fast forward to the present, not only have there been great advancements in this area, the technology as a whole has matured significantly, even more so in recent years. It evolved to the point where it has become less niche. The HDD is far from perfect and protocol aside, it is no different than its ancestor from the 60’s.

Comments are closed.