1 |
On Fri, Jun 05, 2009 at 02:41:33PM -0600, Maxim Wexler wrote |
2 |
|
3 |
> hohndel doesn't specify any options, just says, use ext2. What does |
4 |
> the group recommend? |
5 |
|
6 |
This is in the context of "cheap SSD" diask drives. They behave |
7 |
differently from hard disks, so "everything you know is wrong" if you |
8 |
try stuff that applies to harddrives. You want to minimize the number |
9 |
of writes to disk as much as possible, for 2 reasons... |
10 |
|
11 |
1) Cheap SSD disk drives are atrociously slow. The fewer unnecessary |
12 |
writes, the more responsive the machine will be. |
13 |
|
14 |
2) Cheap SSD disk drives are liable to wear out after a lot of writes. |
15 |
Fewer writes means a longer lifetime for the drive. |
16 |
|
17 |
This brings us to the question of exactly how to minimize writes. I |
18 |
suggest ext2fs with the "async" and "nodiratime" and "noatime" options. |
19 |
|
20 |
* journalling file systems pound away continuously at the journal file. |
21 |
As Martha Stewart would say... "that is not a good thing" for cheap |
22 |
SSD drives |
23 |
|
24 |
* async is the default, but it doesn't hurt to list it. It says not to |
25 |
do a separate write for each request, but to do writes in bigger, and |
26 |
less frequent, batches. The "mount (8)" manpage has this to say about |
27 |
sync... |
28 |
> In case of media with limited number of write cycles (e.g. some flash |
29 |
> drives) "sync" may cause life-cycle shortening. |
30 |
|
31 |
* nodiratime says not to update directory inodes when accessed. You do |
32 |
need to specify it, because it is not the default |
33 |
|
34 |
* noatime says not to update file inodes when accessed. You do need to |
35 |
specify it, because it is not the default. I don't know if "noatime" |
36 |
implies "nodiratime", but I'd play it safe and specify both. |
37 |
|
38 |
-- |
39 |
Walter Dnes <waltdnes@××××××××.org> |