32 GB Flash Storage Drive Announced

Audrius writes to tell us TG Daily is reporting that Samsung has just announced a new 32 GB Flash storage device. The aim of this new solid state disk (SSD) drive is to completely replace the traditional hard drives in many laptops on the market. Some of the advantages offered are the 1.8" form factor, read speeds more than twice that of a normal hard drive, and the promise of 95% less power use.

Comments

What about the limited number of writes?

[jay2003]

I have the understanding that flash memory has a finite number of writes and that conventional filesystems with their update of metadata even on file read could essentially wear out a flash drive quickly if it was used as the main disk drive (as opposed to digital camera use or the like where access is comparably infrequent)

In my experience...

[SheeEttin]

and the promise of 95% less power use

In my experience, promised things usually fall flat on their face. Microsoft springs immediately to mind.

And hopefully, Flash drives will replace the current magnetic platter ones. It's kind of odd for one of the most important devices in a computer to be the only moving one (And therefore the most susceptible to damage, especially in laptops).

I'd buy it

[Eightyford]

I'd buy it. All that is needed is a wireless link to a network attatched file server. 32 GB holds a lot of non-multimedia files.

This is big news.

[zymano]

Only price is the barrier now for the slllloooooooowest parts of a computer.

Re:Data Integrity

[nbert]

Just make sure your filesystem rarely does or needs defragging, and does not log every read.

On a flash drive it's not really important into how many segments a file is split or where they are located since there's no head spinning back and forth. So there's only a problem if your fs does defraging automatically, but it's quite easy to switch this off (at least for developers)
Guess we have to reconsider some habits we've got accustomed to if traditional hds are replaced.

Fragmentation?

[Dr. Evil]

Does fragmentation matter when there are no heads to move?

Re:Fragmentation?

[temojen]

Data-Journalled filesystems (eg Reiser4) keep data in the journal too, not just metadata (those are metadata-journaled filesystems eg Ext3FS), so each block may have parts of several files, and each file may be spread across many more blocks than it would fill on it's own.

You eventually have to consolidate the data of each file. Not nescesarily to sequential blocks, but so files are not sharing blocks.

For flash memory, non-journalled filesystems like Ext2 (mounted -o noatime) may be best. Although that still tries to keep large chunks of files sequential. It might be better to have a non-journalled filesystem that does not pre-allocate inodes and data blocks, but just keeps a free block list and allocates from it in Least-Recently-Used order.

flash wear-out

[soundofthemoon]

Flash memory cells will indeed wear out after some number of writes. This number is typically pretty high, on the order of a million writes. For most file operations that will probably be a higher MTBF than a magnetic disk with moving parts. Any significant problem would be with hot spots, like VM backing store and file system tables. However you can level wear by using cells in a something like a round-robin fashion. Remember that contiguity isn't an issue with flash because there is no seek time waiting for the head to move. There will probably be some challenges in balancing wear leveling against optimizing file system and VM performance, but in the long run flash drives will likely be much faster and more reliable than magnetic disks.

Re:Interesting ....

[cliveholloway]

$300? I just bought one for $160 on eBay. I think it must be a while since you last looked at prices ;-)