全面介紹Linux內核中的NUMA系統，不錯的資料。
1.Introduction
2.A sample NUMA system
3.Linux and NUMA memory
3.1.Memory Management 101
3.2.NUMA memory
4.Efficient allocations
4.1.Optimal placement: Node local
4.2.Multi node applications
4.3.Page Cache
4.4.Memory balancing
4.5.NUMA interlink and node hardware limits
5.Memory reclaim
5.1.Reclaim in an UP or SMP system
5.2.NUMA reclaim
6.Memory Migration
6.1.Why migrate pages
6.2.How page migration works
6.3.Determining how a process uses memory
6.4.Cpuset controlled memory migration
6.5.Policy based page migration via mbind()
6.6.The numa_migrate_pages() system call
6.7.The migratepages command line tool
6.8.Migrating individual pages using the move_pages() system call
7.Memory Policies
8.Cpusets
9.Potential future work
9.1.Conceptual overlap between cpusets and memory policies
9.2.Cpuset interaction with memory subsystem managing their own locality (f.e. Slab)
9.3.Memory policy support for the page cache
9.4.Page dirtying in shared writable mapping not accounted for
9.5.Lock contention when reading from a file in a multithreaded process
9.6.Better memory reclaim for SLABs
9.7.Manual vs. Automatic page migration
9.8.NUMA scheduler with per node statistics
10.Conclusion
11.References
點擊下載：Local and Remote Memory: Memory in a Linux NUMA System
