Friday, November 04, 2005
Interesting article on flash-based HDDs
Was reading an interesting article on flash-based HDDs. It seems they're starting to gain some traction in the high-end enterprise market. Proponents are citing durability (No moving parts, obviously), speed (Higher speed flash is already way faster than HDDs by several times, and even lower end flash has almost nil seek times), and reliability.
It's that last one that interests me. Flash has a limited rewrite count before the chips fail. With HDDs, it's not so well defined. It seems that flash chips are up to about 1 million writes before they fail, which leads to four or five years in servers. Obviously it's way higher in consumer products due to lower disk activity. Still, it's not quite high enough. They're claiming that four or five years is already more reliable than most magnetic HDDs, but I'd give it a few years before I'd feel comfortable with that claim.
Another claim they're making is that flash capacity is going up about 35% per year, and prices down by 35% per year. Looks like moving entirely to flash for storage in a general purpose PC isn't so far off after all. The idea of hybrid drives is already floating around (Using flash as an enormous multi-gig cache on magnetic drives), and that's a bit more likely to happen before full-on flash drives become popular.
There are alreay some flash-based IDE hard drives (In 2.5" notebook drive form factor as a drop-in replacement for a magnetic drive, I'm not talking about a DOM here) on the market, but I don't think they come bigger than 16 gigs or so. They cost something like $100 to $150 per gig anyhow.
Still, I drool about the day when flash drives take over. Flash has significantly higher transfer rates than magnetic drives (with better future potential too), comparatively non-existant seek times, and way lower power consumption. The seek times mean that disk fragmentation becomes effectively meaningless, because there is no penalty for grabbing another fragment from elsewhere on the disk.
I imagine the disk defrag software vendors aren't too happy about that, though :P
The original article can be read here: http://www.tgdaily.com/2005/11/04/ssd_harddrives/
It's that last one that interests me. Flash has a limited rewrite count before the chips fail. With HDDs, it's not so well defined. It seems that flash chips are up to about 1 million writes before they fail, which leads to four or five years in servers. Obviously it's way higher in consumer products due to lower disk activity. Still, it's not quite high enough. They're claiming that four or five years is already more reliable than most magnetic HDDs, but I'd give it a few years before I'd feel comfortable with that claim.
Another claim they're making is that flash capacity is going up about 35% per year, and prices down by 35% per year. Looks like moving entirely to flash for storage in a general purpose PC isn't so far off after all. The idea of hybrid drives is already floating around (Using flash as an enormous multi-gig cache on magnetic drives), and that's a bit more likely to happen before full-on flash drives become popular.
There are alreay some flash-based IDE hard drives (In 2.5" notebook drive form factor as a drop-in replacement for a magnetic drive, I'm not talking about a DOM here) on the market, but I don't think they come bigger than 16 gigs or so. They cost something like $100 to $150 per gig anyhow.
Still, I drool about the day when flash drives take over. Flash has significantly higher transfer rates than magnetic drives (with better future potential too), comparatively non-existant seek times, and way lower power consumption. The seek times mean that disk fragmentation becomes effectively meaningless, because there is no penalty for grabbing another fragment from elsewhere on the disk.
I imagine the disk defrag software vendors aren't too happy about that, though :P
The original article can be read here: http://www.tgdaily.com/2005/11/04/ssd_harddrives/
