Linux core complementing: Performance and danger

Linux core complementing: the performance and danger are at the same time-Linux general technology-Linux programming and kernel information. The following is a detailed description. The patches involved in this article can be found in mm-patches and ck-patches.

Update: due to conflicts inherent in Linux core development,
Some developers will not perform core development,
Some patches mentioned below will not be maintained.

Swap-prefetch:
It is relatively stable. 2.6.22 is expected to enter mainline.
It has already been used by Novell SuSe Desktop 10 core.
There may be no acceleration effect, but it is generally regarded as Yes (it accelerates firefox ).

File:
Mm-implement-swap-prefetching.patch

Brief description:
Implement swap prefetching when the vm is relatively idle and there is free
Ram available.

Prefetching can be enabled/disabled via the tunable in
/Proc/sys/vm/swap_prefetch initially set to 1 (enabled ).

Enabling laptop_mode disables swap prefetching to prevent unnecessary spin
Ups. (The recently added feature saves a lot of power for notebook users)

O (1) scheduler: CFS scheduler v15
Relatively stable. 2.6.23 is expected to enter mainline.

Readahead:
Stable (although the development time is not short ).
Wu fengguang, author of rare Chinese works

The acceleration effect is self-evident. Pre-read files.

File: All files starting with readahead.

Unstable and the effect is obvious. Consider the following:
Big IO commands v6
Large Blocksize Support V3
(These two need to be searched by themselves, not included in mm and ck .)

It took countless times, 2.6.23, and reiser4 again.