So I presented at a VMUG in the UK earlier this week. Topic: The Future of Storage. Cracking topic I thought…. I can make up whatever I want and not get shot down. But more seriously, storage is one of the most interesting areas of enterprise tech at the moment, with so much going on.
Anyway, I talked about three main things:
1. The future of the disk drive
2. Solid State and NVM technologies
3. Software Defined Storage
In this post I’ll cover some of what I talked about in regards to the disk drive. And actually, as boring as the disk drive is these days – I think there’s some pretty significant stuff going on with it. So not boring!
Capacity Monster but Performance Disaster
Anyway, the writing I’m seeing on the wall for the disk drive is all about it’s unstoppable march towards becoming a capacity monster and a performance disaster. And stick with me here. I originally thought I was talking long term, but I’m actually not so sure now….
I figured I’d grab the audience’s attention by slapping “The Petabyte Hard Drive” on the top of the slide. Obviously I don’t think we’ll ever have a PB hard drive, but….. capacity is definitely what the disk drive will be all about in the future. I’m convinced of that.
I look at emerging technologies like Shingled Magnetic Recording (SMR) and Heat Assisted Magnetic Recording (HAMR). SMR will up recording density by a decent amount – something like 20-30%. Great. But it’ll annihilate performance! Annihilate it!!
Read performance on SMR media shouldn’t be too bad, and as far as I can tell, first writes to a track *could* be OK. But performing updates to tracks and sectors that you’ve already written to… well…. the recommendation would be “don’t even bother”. Updates-in-place will be like a visit to a 19th century English dentist!
So if performance will tank so much, what’s the point of technologies like SMR? Well, it seems that spinning media is heading towards the write-once-read-many road. And while that’s not much use to enterprise IT, it’s perfect for the exploding arena of social media and other unstructured web 2.0 content – photos, videos and the likes. After all, we upload photos and videos once and then never update them. And… they’re pretty much sequential read hogs. So ideal for this new generation of disks. And let’s be honest, that’s a huge market that’s growing like mad. And the drive manufacturers go where the money is (growth sectors). Enterprise IT just takes what’s available and makes do!
But what about 10K and 15K drives?
About 3-4 years ago I was in a meeting with a guy from HGST talking about, well….. disk drives. And back then he predicted the end of the 15K drive. Reason? Well two reasons actually. The first reason was flash related, in that the extra effort to manufacture 15K vs 10K wasn’t worth the modest performance gain. Especially when you consider the significant difference in performance between a 10K and flash device. Anyway, the second reason was market forces. The R&D and manufacturing of 10K drive might not be viable as flash and other solid state media becomes more prevalent.
I’m contemplating a future where our three tiers are NVM (PCM, STT-RAM…), NAND flash, and NL-SAS.
The facts are….. perpendicular recording is all but spent, and new innovations are needed to keep the disk drive viable. The growth markets are the ones that will decide the future of the disk drive, and right now they are social and so-called web 2.0… They want capacity capacity capacity. I’m sure they’re not buying many 10K drives.
Also, some pretty amazing NVM technologies are knocking at the door. These could very well relegate flash to tier 2 before it’s even had a chance to bask on the throne of tier 1.
That leaves the 7.2K/5.4K/SMR/HAMR style spinning drive of the future as tier 3.