Intel Continues Optimizing Linux For Optane DC Persistent Memory Servers
With the Linux 5.15 kernel there is a patch to benefit tiered memory systems with a focus on servers having persistent memory. That patch is demoting pages during page reclamation to slower tiers of memory such as Optane DC persistent memory. Intel continues building on that and other persistent memory kernel work for plumbing the kernel with optimized memory placement for these modern servers.
Intel engineer Huang Ying sent out the latest patch series this week that tunes the Linux kernel's NUMA balancing behavior to optimize memory placement for memory tiering systems. These patches are further optimizing the kernel's dealing of pages in the presence of persistent memory while keeping the most important pages in DRAM.
Most importantly is the end result of this work for an Intel Xeon 2P server with Optane DC persistent memory, "The whole patchset improves the benchmark score up to 115.2%. The basic NUMA balancing based optimization solution (patch 2), the hot page selection algorithm (patch 4), and the threshold automatic adjustment algorithms (patch 6) improves the performance or reduce the overhead (promotion MB/s) greatly."
Those interested in these latest refined patches for optimized memory placement for tiered memory systems can see the v9 patches now residing on the kernel mailing list but hopefully won't be too much longer before seeing mainline.
Intel engineer Huang Ying sent out the latest patch series this week that tunes the Linux kernel's NUMA balancing behavior to optimize memory placement for memory tiering systems. These patches are further optimizing the kernel's dealing of pages in the presence of persistent memory while keeping the most important pages in DRAM.
Most importantly is the end result of this work for an Intel Xeon 2P server with Optane DC persistent memory, "The whole patchset improves the benchmark score up to 115.2%. The basic NUMA balancing based optimization solution (patch 2), the hot page selection algorithm (patch 4), and the threshold automatic adjustment algorithms (patch 6) improves the performance or reduce the overhead (promotion MB/s) greatly."
Those interested in these latest refined patches for optimized memory placement for tiered memory systems can see the v9 patches now residing on the kernel mailing list but hopefully won't be too much longer before seeing mainline.
1 Comment