1 |
Alex Schuster <wonko@×××××××××.org> [10-09-12 04:13]: |
2 |
> meino.cramer@×××.de writes: |
3 |
> |
4 |
> > I think there is some misunderstanding: |
5 |
> > |
6 |
> > Before migration to 64bit: |
7 |
> > |
8 |
> > /dev/sda3 is mounted on / and contains the 32bit Gentoo |
9 |
> > |
10 |
> > /dev/sda10 is mounted on /home/mcc/migration and will contain the |
11 |
> > stuff of the 64bit Gentoo |
12 |
> > |
13 |
> > After migration I will *not* mount /dev/sda10 on / but will clear all |
14 |
> > stuff from /dev/sda3 and move the contents from /dev/sda10 to |
15 |
> > /dev/sda3. |
16 |
> > |
17 |
> > Is still valid what you said under this premissions, Wonko? |
18 |
> |
19 |
> That's how I understood it, although I assumed the temproary 64bit install |
20 |
> would be on a 2nd drive, thus you would copy it back once it seems to |
21 |
> work. No, I see no problem with this. |
22 |
> |
23 |
> About performance: I'm not sure it will be even noticeable. Yes, most |
24 |
> drives (but not all) are organized so the first partitions go to the |
25 |
> outside, which is faster. With LVM, I used to create two volume groups on |
26 |
> my drive, a group for swap and the system, and another one for data. But |
27 |
> then I thought it's not worth the effort, and I lose some of the LVM |
28 |
> benefits. Well, with everything encrypted I don't get full performance |
29 |
> anyway, so my case might be a little different. |
30 |
> |
31 |
> But the performance increase is only true when reading lots of data. I'm |
32 |
> not sure how big the role of this is in real life. Access time is not |
33 |
> influenced, it will on average take half a turn of the drive till the |
34 |
> heads can access the data, and to me it looks like typical stuff a linux |
35 |
> system does is reading many not so large files, cluttered around in the |
36 |
> file system. But that's my guess only. And I understand that you like to |
37 |
> optimize stuff - I like to do this too. But sometimes I think that the |
38 |
> potential benefit might not be so large, compared to the time I spend |
39 |
> moving data around to the ideal place, or the time I would need to spend |
40 |
> thinking about how to tune things. Or the time you need to fix a problem |
41 |
> that you know was working in the old system, but this is gone now and you |
42 |
> cannot have a quick look at it, or just boot into it. You lose the |
43 |
> opportunity to start your old system in order to compare the times of your |
44 |
> big renderings. And maybe at one point you need to create some true 32bit |
45 |
> applications? Happened to me. So I just chroot into my old system and |
46 |
> build there. |
47 |
> |
48 |
> Oh, and you mentioned databases. Yes, mysql stores itsa data in machine- |
49 |
> depenent form. You will need to dump the data and re-import it in the new |
50 |
> system. You will be happy to still have the 32bit system in such a case :) |
51 |
> |
52 |
> Wonko |
53 |
> |
54 |
|
55 |
I also cannot evaluate the real impact the position of the /-partition |
56 |
on the harddisk has on system performance. I read about it years ago |
57 |
and since than I always put the partitions always in the sequence of |
58 |
"boot","swap","root","home" onto the harddisks. May be its only a |
59 |
tradition nowadays... ;) |
60 |
|
61 |
Do you know of any other kind of data beside databses, which may be |
62 |
machinedependant or cause trouble while migrating to 64bit? |
63 |
|
64 |
Best regards, |
65 |
mcc |