1 |
> I've had somewhat negative results with WOLK. On a Dell SMP box with |
2 |
> hardware RAID, it did not work very well at all. (In contrast, |
3 |
> development-sources has worked perfectly.) On two other Dell single CPU |
4 |
> boxes, it ran fine for several months and then both machines crashed |
5 |
> yesterday, so I'm probably going back to vanilla-sources as soon as I |
6 |
> can. |
7 |
|
8 |
In terms of the overall quality of the codebase, I think 2.6 eclipses 2.4 |
9 |
and will continue to mature at a sterling rate. 2.6.0 has to have been the |
10 |
best point "0" release ever. It's going to take a while for peripheral |
11 |
efforts to migrate to the land of 2.6 though. |
12 |
|
13 |
With regard to 2.4, aa-sources is now far and away my favourite kernel |
14 |
patchset and is particularly good for servers, in my view. Obvious |
15 |
highlights are the very clean O(1) scheduler implementation, reduced |
16 |
latency, and numerous cleanups for race conditions and SMP scaleability |
17 |
enhancements. Also, I believe that -aa currently remains the best choice |
18 |
for systems which very large amounts of memory and/or subjected to high |
19 |
levels of memory stress (actually, the most outstanding VM patches from |
20 |
Andrea were integrated mainline in 2.4.23-pre4 - and should have been |
21 |
considerably longer beforehand - but there are still a few extras in -aa). |
22 |
I ran it for some time before taking the plunge apropos of 2.6 "testing" |
23 |
and its performance and stability was exemplary. My faith in this kernel |
24 |
is also based on seeing just how many "3rd party" patches Andrea had |
25 |
decided to include in the patchset for so long, so many of which were |
26 |
recently accepted by Marcelo. |
27 |
|
28 |
Having said that, if any doubt I think vanilla is now a solid choice, by |
29 |
which I mean that the overall quality has improved massively as of 2.4.23, |
30 |
and even more so through 2.4.24 and the latest snapshots. With earlier |
31 |
versions, I would have felt obliged to hunt around for patches to address |
32 |
various issues. |
33 |
|
34 |
--Kerin Francis Millar |