Linux – Is Flash Drive Wear a Significant Issue?

filesystemshard drivelinux

My netbook has a flash drive instead of a hard disk drive, and I'm using Ubuntu Netbook Remix with ext3 as the file system. I've read some articles concerning flash drive wear, and the main concerns seem to be:

The amount of write cycles – each cell can be written to only a limited amount of times (Wikipedia has numbers ranging from 1,000 to 100,000)
You can only write data on a "sector" once, and after that the whole block needs to be erased to use again – and these blocks are ranging from 16 KB to 128 KB.

These are said to add up so that normal file systems that aren't designed to take this into account end up using wearing out the flash drive by moving small amounts of data.

Now I don't doubt that the problem is theoretically very real. However, I know we tech people get easily carried away by interesting optimization problems, such as designing an alternative file system to combat flash wear. For example it's great to do memory optimization, but if you end up saving 100 KB of memory when there's hundreds of MB available anyway, it's not fixing a real problem.

What I end up getting from all this, is that I shouldn't use normal file systems on flash drives because they quickly eat up the drive. But I'm not convinced. So the question is: Is flash drive wear actually relevant in everyday, normal usage? Is my laptop, using ext3, going to eat up my flash drive in few years… or is all of this rather a theoretical problem that does reduce the usage time, but only by so little that it'll never happen in normal conditions? Or is transparent, hardware wear leveling already being used on netbook flash drives to fix the problem, so that an alternative file system wouldn't even do any good?

Sources: (1), (2), (3), (4), (5), (6)

Best Answer

Generally speaking, Flash drive wear and tear is always brought up (SSD and USB) However, I haven't seen it.

I have personally found that cheap USB Flash drives for example go faulty and simply do not get recognised well before you actually see wear and tear.

Also, newer drives use technologies that randomise the locations of writes. I suppose, lets say you have a 100 GB drive and fill it up with 99.5 GB's, then you keep using .5 GB over and over again, you can reach the limit, but again, I use SSDs and USB sticks on a daily basis for very heavy use (over the past few years) and generally speaking, the drives die of general failures well before you see this as a problem.

No Experience with alternative File Systems, However I personally wouldn't bother... Use a mature file system and if it fails within a usable time, take it back under warranty. (if in the UK, up to ~6 years under the sale of goods act as you can say it was designed with a fault and not fit for the purpose of storing data... I am not a lawyer, but I took a laptop back 4 years after buying for a similar reason).

Also, for Windows just maybe worth a look in, I remember seeing a product from Diskeeper, that looks interesting - meant to optimise and extend the life of SSD disks, but I am wondering if it is needed and found several articles doubting it (only linked to one) and goes in to detail about wear and tear. Also, I can not longer see the product on their website, so it must of either been scrapped or built into a different edition.

JFFS2

Includes compression and elegant wear leveling protection.

YAFFS2

Single thing that makes the difference: short mount times, after successful umount.
Implements write once property: once data is written to one block, there is no need to rewrite it. This is important, as it reduces wear.

LogFS

Not very mature, but already included in Linux kernel tree.
Supports larger filesystems than JFFS2/YAFFS2 without problems.

UBIFS

More mature than LogFS
Write caching support
On scalability: article. On large disks, better performance than with JFFS2

ext4

If no driver or card (for example SSD drives do have internal wear leveling, at least usually) handle wear leveling, then ext4 is not the best idea, as it is not intended for raw flash usage.

Which one is the best?

Of course, it depends on usage and support. From what I read on the Internet, I would recommend UBIFS. Good support for large filesystems, mature development phase, adequate performance and no huge downsides.

Linux – How to correctly partition usb flash drive and which filesystem to choose considering wear leveling

Don't worry about cylinders, heads and sectors and only ever work with sectors. The easiest way to align the filesystem is to simply not have a partition table (i.e. create the filesystem on /dev/sdX directly). Most Linux systems will handle this fine, but Windows will freak out and you also won't be able to make it bootable. If you need a partition table, work out how many 512 byte sectors make up an erase block. Erase blocks are around 128-512KiB so if you can't find out what your flash drive uses go with 1024 sectors. Make sure that partitions start on a multiple of this number.

You also want to tell ext2/3/4 about the erase block size so it can avoid unnecessarily modifying blocks. Set both stride and stripe-width to the number of filesystem blocks which make up an erase block, so working on a 512KiB erase block use the following:

mkfs.ext4 -b 4096 -E stride=128,stripe-width=128

In terms of performance, not having a journal should improve performance (but increase the chance of data corruption if operation is interrupted). However, ext4's extents should improve performance because less metadata needs to be modified for large files. I would probably use ext4 with the journal disabled:

tune2fs -O ^has_journal