1 |
On Saturday 26 March 2011 15:06:31 Elaine C. Sharpe wrote: |
2 |
> > Just because something works for most people, doesn't mean it will for |
3 |
> > everyone either. If you lose data, it doesn't matter. LVM just adds |
4 |
> > one more layer of something to go wrong. Me, I don't need the extra |
5 |
> > risk of having a system that doesn't boot and a loss of data. I'm sure |
6 |
> > there are a lot of people that see it the way I do too. They just |
7 |
> > don't |
8 |
> > need the extra risk. |
9 |
> |
10 |
> Using the least number of layers of abstraction you can get away with is |
11 |
> a perfectly valid criteria. What I was pointing out was that informal |
12 |
> polls of users with a sad story to tell is not a very effective way to |
13 |
> conduct research. People say all kinds of things that just aren't true. |
14 |
|
15 |
There's an elephant in this room. The number of actual layers is greater than |
16 |
just LVM plus FS. It's whatever the BIOS (or a reasonable substitute is |
17 |
doing), plus the drive firmware, kernel driver(s) - there's more than one of |
18 |
those - plus any RAID in use (hardware or software) and finally the file |
19 |
system. |
20 |
|
21 |
That's a lot of layers, a lot of code, a lot of opportunity for people to |
22 |
reveal the extent of their lack of knowledge. I've often heard it said that |
23 |
code like ZFS and brtfs eliminates several of these layers therefore it's |
24 |
technically a better option. That may be true, but let me just point out that |
25 |
whatever LVM+fs+other_stuff is doing as separate chunks of code also gets done |
26 |
by ZFS etc. You just don't see it, and just because it's abstracted away |
27 |
doesn't mean it's not there. |
28 |
|
29 |
-- |
30 |
alan dot mckinnon at gmail dot com |