We appreciate feedback about this patch series from the community. Historically, neither the performance nor latency of the swap path mattered. The underlying I/O was slow enough to hide any latency coming from software and the low IOPS kept the overall CPU impact low.
Times have changed. Coming generation of Solid state Block device latencies are getting down to sub 100 usec, which is within an order of magnitude of DRAM, and their performance is orders of magnitude higher than the single- spindle rotational media we've swapped to historically. This could benefit many usage scenearios. For example cloud providers who overcommit their memory (as VM don't use all the memory provisioned). Having a fast swap will allow them to be more aggressive in memory overcommit and fit more VMs to a platform. In our testing [see footnote], the median latency that the kernel adds to a page fault is 15 usec, which comes quite close to the amount that will be contributed by the underlying I/O devices. The software latency comes mostly from contentions on the locks protecting the radix tree of the swap cache and also the locks protecting the individual swap devices. The lock contentions already consumed 35% of cpu cycles in our test. In the very near future, software latency will become the bottleneck to swap performnace as block device I/O latency gets within the shouting distance of DRAM speed. This patch set, plus a previous patch Ying already posted (commit: f6498b3f) reduced the median page fault latency from 15 usec to 4 usec (375% reduction) for DRAM based pmem block device. Patch 1 is a clean up patch. Patch 2 creates a lock per cluster, this gives us a more fine graind lock that can be used for accessing swap_map, and not lock the whole swap device Patch 3 splits the swap cache radix tree into 64MB chunks, reducing the rate that we have to contende for the radix tree. Patch 4 eliminates unnecessary page allocation for read ahead. Patch 5-8 create a per cpu cache of the swap slots, so we don't have to contend on the swap device to get a swap slot or to release a swap slot. And we allocate and release the swap slots in batches for better efficiency. Ying Huang & Tim Chen Footnote: Testing was done on 4.8-rc3-mm1 kernel with/without optimizations from this patche series plus one additional patch Ying posted earlier on removing radix tree write back tag in swap cache. Eight threads performed random memory access on a 2 socket Haswell using swap mounted on RAM based PMEM block device. This emulated a moderate load and a SWAP device unbounded by I/O speed. The aggregate working set is twice the RAM size. We instrumented the kernel to measure the page fault latency. Change Log: v2: 1. Fix bug in the index limit used in scan_swap_map_try_ssd_cluster when searching for empty slots in cluster. 2. Fix bug in swap off that incorrectly determines if we still have swap devices left. 3. Port patches to mmotm-2016-10-11-15-46 branch Huang, Ying (3): mm/swap: Fix kernel message in swap_info_get() mm/swap: Add cluster lock mm/swap: Split swap cache into 64MB trunks Tim Chen (5): mm/swap: skip read ahead for unreferenced swap slots mm/swap: Allocate swap slots in batches mm/swap: Free swap slots in batch mm/swap: Add cache for swap slots allocation mm/swap: Enable swap slots cache usage include/linux/swap.h | 35 ++- include/linux/swap_slots.h | 37 +++ mm/Makefile | 2 +- mm/swap.c | 6 - mm/swap_slots.c | 306 +++++++++++++++++++++++++ mm/swap_state.c | 76 +++++- mm/swapfile.c | 558 +++++++++++++++++++++++++++++++++++---------- 7 files changed, 875 insertions(+), 145 deletions(-) create mode 100644 include/linux/swap_slots.h create mode 100644 mm/swap_slots.c -- 2.5.5