| 1 |
On Thu, Sep 18, 2014 at 12:27 PM, Kerin Millar <kerframil@×××××××××××.uk> wrote: |
| 2 |
> |
| 3 |
> The need for the OOM killer stems from the fact that memory can be |
| 4 |
> overcommitted. These articles may prove informative: |
| 5 |
> |
| 6 |
|
| 7 |
A big problem with Linux along these fronts is that we don't really |
| 8 |
have good mechanisms for prioritizing memory use. You can set hard |
| 9 |
limits of course, which aren't flexible, but otherwise software is |
| 10 |
trusted to just guess how much RAM it should use. |
| 11 |
|
| 12 |
It would be nice if processes could allocate cache RAM, which could be |
| 13 |
preferentially freed if the kernel deems necessary. If some pages are |
| 14 |
easier to regenerate than to swap, this could also be flagged (I have |
| 15 |
a 50Mbps connection - I'd rather see my browser re-fetch pages than go |
| 16 |
to disk when the disk is already busy). There are probably a lot of |
| 17 |
other ways that memory use could be optimized with hinting. |
| 18 |
|
| 19 |
-- |
| 20 |
Rich |
Archives