> Implementing page colouring into the underlying buddy page allocator
> was tried, but given up because it actually slowed down some things and
> leads to bad fragmentation [it is probably required to replace the simply
> and fast buddy allocator with a more sophisticated zone allocator to
> prevent fragmentation].
Now that we speak of it, does anybody have some pointers
to documents about zone allocators, or does someone have
some nice & simple algoritms in mind?
Some of the things that come to mind:
- use 128K (or 512K) zones
- allocate non-swappable and swappable stuff in separate zones
- allocate 4-page and bigger stuff in a separate zone from 1-page
allocs
- use page coloring on >16M machines, we allocate 8 zones for
user-stuff at a time and switch zones after every allocation
- inside zones, we could use a buddy-like scheme, but that
probably isn't needed
- we fill up each zone as much as possible before allocating
a new (batch of) zone(s)
- there's a large penalty involved when kswapd encounters a
zone with just a single swappable page inside it (when there
aren't enough empty zones available)
- ??? what else ???
Rik.
+-------------------------------------------+--------------------------+
| Linux: - LinuxHQ MM-patches page | Scouting webmaster |
| - kswapd ask-him & complain-to guy | Vries cubscout leader |
| http://www.phys.uu.nl/~riel/ | <H.H.vanRiel@phys.uu.nl> |
+-------------------------------------------+--------------------------+
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu