Tags · fio-prf/em

em-20121130

swap: make cluster allocator per-cpu

swap cluster allocatior is to get better request merge to improve performance.
But the cluster is shared globally, if multiple tasks are doing swap, this will
cause interleave disk access. While multiple tasks swap is quite common, for
example, each numa node has a kswapd thread doing swap.

We makes the cluster allocator per-cpu here. The interleave disk access issue
goes away. All tasks will do sequential swap.

In my test (swap to a 2-disk raid0 partition), this improves around 10%
throughput, and request size is increased significantly.

This might improve swap readahead too. Considering page reclaim always swaps
several adjancent pages, this will make swap readahead not mis-prefetch data
for sequential access pattern workload. For random access pattern workload,
there should be no overhead as we only enable the per-cpu allocator for SSD, so
no seek penality.

Signed-off-by: Shaohua Li <shli@fusionio.com>