Thursday, December 31, 2009

SSDs And Their March Forward

I was lying in bed reading an article this morning on SSD technology and the economics of this compared to HDD and the author's comments on when and why we'll move from HDD to SSD.  I pretty much agreed with the article, however one area that people don't seem to have caught on when it comes to SSDs is that although everyone says "HDD will replace Tape for backups" it simply hasn't in the enterprise market and it won't - however SSD will.

This is the coment I made on that article:

Aside from the impact NAND SSDs are starting to have and will increasingly have in the hard drive space, one big area people continually seem unable to focus on, nor even recognise, is that they will be a disruptive technology when it comes to the backup and archive market which is currently dominated by tape drives and media.

People keep claiming that hard drive backups will replace tape backups and they've made nothing more than a dint in the backup market and barely a scratch in the archive market. Sure, Microsoft may have dropped tape support in their Windows Server 2008 operating system's native backup application, but pretty much anyone who uses tape drives for backup uses both a third party tape driver and third party backup application anyway. This change by Microsoft will have very little impact on the devices used for backups in the IT industry.

What *will* make an impact, in my opinion, is NAND SSD technology.

HDDs are too delicate to be taken seriously as a backup and archival technology in the enterprise arena. Sure, they are a cheap and viable option in the SMB market, but tape still has enough advantages down here to be a valid contender. However Flash will become a serious contender as capacities increase. One of the reasons was touched on in your article - IOPS. As the backup window decreases, the slow (relative) speed of tapes means that they can't be used directly for backups and an initial backup is made to HDD on either that server itself (on an additional spindle set) or to another dedicated backup server, then this is taken to tape. All of this adds cost and another layer of complexity which is, quite effectively, more cost.

When SSD capacity and cost starts coming into competition with tape drives (and this will happen before SSDs compete with HDDs due to the expense of tape drive systems), we'll find enterprises starting to look seriously at SSD as a cost effective replacement for tape for both backup and archival purposes.

Another reason for SSD being extremely useful for backups is the speed of restore which will vary from significantly faster to a few orders of magnitude faster.

So, both technically and economically, SSDs will play an increasingly important role as not only primary storage, but also as backup and archival storage in place of the Tape.
Then I got out of bed, put the kettle on for my first cuppa and posted this blog post.  :)


The Outspoken Wookie

(Trying to remember that he's on holidays.  As in *NOT WORK*.  Holidays.  He has 10-15 minutes work to do today and that's it, so why is he thinking about this stuff this early in the morning?)


Chris Knight said...

SSDs still have electronic circuitry, while tapes are still (mostly) mechanical. This means that tape still has a greater longevity than HDDs/SSDs.

The other vitally important factor is that SSDs are essentially running their own proprietary file systems to emulate a hard disk. This makes data recovery a more difficult proposition in case of corruption caused by circuitry failure. I think we'll need to see a standard - even if it's an industry one - for the SSD file systems to aid in the SSD data recovery space and uptake of SSDs as a short term archival media.

IOPS can be overcome by using RAIT, but then you're also increasing the number of tapes + drives required for a backup.

Mechanical devices still trump electronic ones when it comes to archive and recovery capabilities.

It's an interesting dilemma given the cost, backup set size, archival duration, backup window and recovery capability requirements.

Hilton Travis said...

Yeah, but the number of tape drives that eventually go far enough out of alignment resulting in tapes that are not readable on new drives isn't insignificant either. The issue with an industry-standard SSD format is that it would need to support NTFS and all other filesystem security and other features, which is something that doesn't exist today and won't likely exist in the future due to the various different licensing, IP and plain backstabbing that goes on between MS and pretty much everyone else.

IOPS definitely can be overcome by RAID, but how many businesses employ RAID-5 tape drives?

Mechanical devices still (as in currently) trump electronic ones. I agree. But I was talking about the future in this post and in the future, I can see electronic devices (SSDs or similar) trumping tapes in many areas.