1 |
On Sun, 30 May 2010 14:20:36 +0000 (UTC) |
2 |
Grant Edwards wrote: |
3 |
|
4 |
> On 2010-05-30, Neil Bothwick <neil@××××××××××.uk> wrote: |
5 |
> > On Sat, 29 May 2010 07:59:31 -0400, David Relson wrote: |
6 |
> > |
7 |
> >> Indeed flash drives _do_ have a lifetime. My recollection is that |
8 |
> >> it's in the thousands of writes if not the hundreds of thousands |
9 |
> >> of writes. Assuming a life of 1,000 writes and you backup once |
10 |
> >> daily, that's 3 years of backups. 10,000 writes would be 30 |
11 |
> >> years. Of course if you backup every hour, 10,000 writes is a |
12 |
> >> year (or so). |
13 |
> > |
14 |
> > You're assuming that each backup only writes once, which is far from |
15 |
> > true. If you mount a drive with the sync option, the FAT is updated |
16 |
> > for every block you write, so even a single file can cause |
17 |
> > thousands of writes to the same location. |
18 |
> |
19 |
> And you're assuming that the flash controller chip in the USB drive |
20 |
> doesn't do wear-leavelling. |
21 |
|
22 |
FWIW, I have enabled synchronous writes for a Disk-On-Module (SSD) |
23 |
formatted ext2. It makes writing take significantly longer and I have |
24 |
had a DOM go bad (become unusable). Admittedly, I don't know whether |
25 |
the DOM does wear-levelling and I don't know the underlying cause of |
26 |
the failure. In any case it was "Not Good (tm)" ... |