Messages in this thread | | | From | Peter Zijlstra <> | Subject | [PATCH 00/34] mm: Page Replacement Policy Framework | Date | Wed, 22 Mar 2006 23:31:40 +0100 |
| |
This patch-set introduces a page replacement policy framework and 4 new experimental policies.
The page replacement algorithm determines which pages to swap out. The current algorithm has some problems that are increasingly noticable, even on desktop workloads. As said, this patch-set introduces 4 new algorithms.
Patches 01 - 25:
Introduction of the general framework. Piece by piece the current use-once LRU-2Q policy is isolated. With each patch a piece of the framework API is introduced.
Patches 26 - 29:
Adds a policy based on CLOCKPro. (http://linux-mm.org/PeterZClockPro2)
Patches 30 - 32:
Adds a policy based on CART. (http://linux-mm.org/PeterZCart)
Patch 33:
Adds a variation of the CART policy that tries to incorporate cyclic access patterns.
Patch 34:
Adds a random page replacement policy, simple policy that uses a simple PRNG to take the decision. More of a toy example than a real alternative.
Individual patches and a rollup can be found here: http://programming.kicks-ass.net/kernel-patches/page-replace/
Measurements:
(Walltime, so lower is better)
cyclic-anon ; Cyclic access pattern with anonymous memory. (http://programming.kicks-ass.net/benchmarks/cyclic-anon.c)
2.6.16-rc6 14:28 2.6.16-rc6-useonce 15:11 2.6.16-rc6-clockpro 10:51 2.6.16-rc6-cart 8:55 2.6.16-rc6-random 1:09:50
cyclic-file ; Cyclic access pattern with file backed memory. (http://programming.kicks-ass.net/benchmarks/cyclic-file.c)
2.6.16-rc6 11:24 2.6.16-rc6-clockpro 8:14 2.6.16-rc6-cart 8:09
webtrace ; Replay of an IO trace from the Umass trace repository (http://programming.kicks-ass.net/benchmarks/spc/)
2.6.16-rc6 8:27 2.6.16-rc6-useonce 8:24 2.6.16-rc6-clockpro 10:23 2.6.16-rc6-cart 15:30 2.6.16-rc6-random 15:52
mdb-bench ; Low frequency benchmark. (http://linux-mm.org/PageReplacementTesting)
2.6.16-rc6 4:20:44 2.6.16-rc6 (mlock) 3:52:15 2.6.16-rc6-useonce 4:20:59 2.6.16-rc6-clockpro 3:56:17 2.6.16-rc6-cart 4:11:54 2.6.16-rc6-random 5:21:30
(I should do more runs to get error bounds on these values, this is the avg of 3)
Aside from tweaking the policies, the big thing left is NUMA-ify the nonresident page trackers.
The results merit further attention, please consider for 2.6.18.
Peter
---
Documentation/vm/page_replacement_api.txt | 216 +++++++ fs/cifs/file.c | 4 fs/exec.c | 4 fs/mpage.c | 5 fs/ntfs/file.c | 4 fs/ramfs/file-nommu.c | 2 include/linux/mm_cart_data.h | 39 + include/linux/mm_cart_policy.h | 141 ++++ include/linux/mm_clockpro_data.h | 21 include/linux/mm_clockpro_policy.h | 143 +++++ include/linux/mm_inline.h | 39 - include/linux/mm_page_replace.h | 146 +++++ include/linux/mm_page_replace_data.h | 19 include/linux/mm_random_data.h | 9 include/linux/mm_random_policy.h | 47 + include/linux/mm_use_once_data.h | 16 include/linux/mm_use_once_policy.h | 175 ++++++ include/linux/mmzone.h | 8 include/linux/nonresident-cart.h | 34 + include/linux/nonresident.h | 12 include/linux/page-flags.h | 11 include/linux/pagevec.h | 8 include/linux/percpu.h | 5 include/linux/rmap.h | 4 include/linux/swap.h | 10 init/main.c | 2 mm/Kconfig | 32 + mm/Makefile | 6 mm/cart.c | 723 +++++++++++++++++++++++++ mm/clockpro.c | 855 ++++++++++++++++++++++++++++++ mm/filemap.c | 20 mm/hugetlb.c | 5 mm/memory.c | 42 + mm/mempolicy.c | 13 mm/mmap.c | 5 mm/nonresident-cart.c | 362 ++++++++++++ mm/nonresident.c | 167 +++++ mm/page_alloc.c | 76 -- mm/random_policy.c | 292 ++++++++++ mm/readahead.c | 9 mm/rmap.c | 26 mm/shmem.c | 2 mm/swap.c | 206 +------ mm/swap_state.c | 6 mm/swapfile.c | 17 mm/useonce.c | 489 +++++++++++++++++ mm/vmscan.c | 552 +++---------------- 47 files changed, 4226 insertions(+), 803 deletions(-)
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |