1 |
On Fri, 2002-09-20 at 01:02, Tantive wrote: |
2 |
> Hi! |
3 |
> |
4 |
> |
5 |
> as I'm the OpenMosix (www.openmosix.org) guy for gentoo I would inform |
6 |
> you about the benefits you could have using openmosix and would ask you |
7 |
> to test the openmosix-ebuilds. |
8 |
> |
9 |
> In the portage tree we have at the moment: |
10 |
> |
11 |
> - openmosix-sources (2.4.18-r5 and 2.4.19-r5 are the latest, but masked) |
12 |
> The patched vanilla-sources including openmosix and evms. |
13 |
> |
14 |
> - openmosix-user (latest is 0.2.4, masked, too) |
15 |
> The userland tools needed to manage your cluster. |
16 |
> |
17 |
> - openmosixview (1.2, guess what... masked) |
18 |
> A nice gui which shows you the current load in you cluster. |
19 |
> |
20 |
> |
21 |
> |
22 |
> OpenMosix will allow you to share your CPU-power across several machines |
23 |
> (x86-only at the moment) building a cluster containing several nodes. |
24 |
> |
25 |
> So let's make an example: You have a slow machine and a fast one. If you |
26 |
> want to compile a new kernel on the slow one OpenMosix will "migrate" |
27 |
> these processes to the fast one. This means you could compile your |
28 |
> kernel at approx. the same speed you would on your good machine. |
29 |
> Having many nodes in your cluster the speed increases with every node. |
30 |
> BUT this happens completely transparent. You will have to do nothing. |
31 |
> Nodes can even join and leave a running cluster with no bad effects. |
32 |
|
33 |
Correct me if I'm wrong, but IIRC, the original mosix was useless for |
34 |
the compilation speed improvements due to the fact that a typical |
35 |
compilation process doesn't last long enough to even be considered |
36 |
for migration. And even if the migration would be forced, the migration |
37 |
overhead compared to the process execution time, would kill all the time |
38 |
savings. |
39 |
|
40 |
Vitaly |