1 |
On Wed, 13 May 2009 01:58:18 +0200 |
2 |
Raph <gibboris@×××××.com> wrote: |
3 |
|
4 |
> Of course, and so sadly, there is not a penny... |
5 |
> In order to make a network where any user can access |
6 |
> his datas whatever the computer he uses, I'm thinking about |
7 |
> the following kind of configuration : |
8 |
> - Each client (one of the 20 computers) runs a MBD daemon |
9 |
> which shares its HD to the server (the 21st) over the |
10 |
> ethernet. |
11 |
> - The "server" handle the 3.2TB in a volume group and |
12 |
> create the logical volumes with LVM |
13 |
> - The "server" also run a NFS daemon |
14 |
> - Then each client can grab it's data through NFS. |
15 |
> - happiness ... ? |
16 |
> ------ |
17 |
> I never did such a setup but already see a couple of |
18 |
> problems : |
19 |
> - each client need some private HD space to, at least, |
20 |
> boot enough to let the NBD daemon start. |
21 |
> - $10 question : if 20 users login, will the ethernet |
22 |
> be fast enough to support the load ? |
23 |
|
24 |
While I've never had any chance to try distributed fs outside the |
25 |
sandbox, it looks like a best solution in your case, providing both |
26 |
speed and data redundancy. |
27 |
|
28 |
Specifically, I mean AFS (there's a bit outdated guide on gentoo-wiki), |
29 |
lustre and several other less-known projects like tahoe or pohmelfs. |
30 |
|
31 |
http://allmydata.org/trac/tahoe |
32 |
http://tservice.net.ru/~s0mbre/old/?section=projects&item=pohmelfs |
33 |
|
34 |
-- |
35 |
Mike Kazantsev // fraggod.net |