1 |
On Sonntag 08 November 2009, Stroller wrote: |
2 |
> On 8 Nov 2009, at 00:10, Volker Armin Hemmann wrote: |
3 |
> > ... |
4 |
> > using a livecds kernel is probably the worst decision out there. |
5 |
> > |
6 |
> > http://www.kroah.com/lkn/ |
7 |
> > |
8 |
> > as you can see, you don't have to download it. |
9 |
> > |
10 |
> > Or just do it step for step, reading help files. |
11 |
> > |
12 |
> > seccomp? Except Andrea Arcangeli nobody uses it. Can be deactivated. |
13 |
> > I2O? |
14 |
> > Almost nobody uses it. Especially not 'commodity' hardware, out it |
15 |
> > goes. Numa? |
16 |
> > Do you have a multi-socket system? No? Then you don't need it. ... |
17 |
> > you can |
18 |
> > remove a lot of cruft that way. Namespaces - you don't need it? Kick |
19 |
> > 'em out. |
20 |
> > Group scheduling? Sure, a great way to reduce performance... |
21 |
> |
22 |
> But Volker, if it takes me an hour to decrapify my kernel config and |
23 |
> make it faster, it will probably take 1000 years for those speed |
24 |
> improvements to pay off. |
25 |
> |
26 |
> If I had unlimited time then I would love to read that book. I really |
27 |
> LIKE the idea of decrapifying my kernel config. But realistically, any |
28 |
> time I spend on it is time wasted, for which no difference will be |
29 |
> appreciable. |
30 |
> |
31 |
> Stroller. |
32 |
> |
33 |
|
34 |
I am not you, but I need maybe 5min for a config ;) |
35 |
|
36 |
and there are more benefits. Smaller binary, more cpu cache free for real data. |
37 |
Better performance lies that way. Also, you don't have to wonder about |
38 |
processes you did not start. Security is also a point. A smaller codebase in |
39 |
use is a saver codebase in use. A lot of bugs only affect kernels with certain |
40 |
features turned on - it is very relaxing if you don't have that feature... |