From f7295a1c4cd07c393ced70ad0c8622efdb6ab26d Mon Sep 17 00:00:00 2001 From: Jakub Wartak Date: Wed, 16 Apr 2025 10:23:31 +0200 Subject: [PATCH v1] Add capability to interleave shared memory across multiple NUMA nodes Introduce new GUC numa=off(default)/on/auto that might be used to enable interleaving shared memory. Until today, imbalances in shared memory allocations on NUMA setups, may have caused non-deterministic performance due to differences in latencies and bandwidths across interconnects ("remote" access). This is only supported on Linux with libnuma. Author: Jakub Wartak Inspired-by: Andres Freund Reviewed-by: Discussion: --- doc/src/sgml/config.sgml | 16 ++++++++++++++++ src/backend/port/sysv_shmem.c | 7 +++++++ src/backend/utils/misc/guc_tables.c | 18 ++++++++++++++++++ src/backend/utils/misc/postgresql.conf.sample | 2 ++ src/include/port/pg_numa.h | 1 + src/include/storage/pg_shmem.h | 10 ++++++++++ src/port/pg_numa.c | 17 +++++++++++++++++ 7 files changed, 71 insertions(+) diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index c1674c22cb2..15397df71d6 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -2313,6 +2313,22 @@ include_dir 'conf.d' + + numa (enum) + + numa configuration parameter + + + + + Specifies wheter to use NUMA interleaving policy for the shared memory + segment. Possible values are off, on + and auto. This parameter is only effective on Linux. + The default value is off. This parameter can only be set + at server start. + + + diff --git a/src/backend/port/sysv_shmem.c b/src/backend/port/sysv_shmem.c index 197926d44f6..510b0e53638 100644 --- a/src/backend/port/sysv_shmem.c +++ b/src/backend/port/sysv_shmem.c @@ -29,6 +29,7 @@ #include "miscadmin.h" #include "port/pg_bitutils.h" +#include "port/pg_numa.h" #include "portability/mem.h" #include "storage/dsm.h" #include "storage/fd.h" @@ -663,6 +664,12 @@ CreateAnonymousSegment(Size *size) allocsize) : 0)); } + if (numa == NUMA_ON || (numa == NUMA_AUTO && pg_numa_get_max_node() > 1)) + { + elog(DEBUG1, "enabling NUMA shm interleaving"); + pg_numa_interleave_memptr(ptr, allocsize); + } + *size = allocsize; return ptr; } diff --git a/src/backend/utils/misc/guc_tables.c b/src/backend/utils/misc/guc_tables.c index 60b12446a1c..e4c9491df78 100644 --- a/src/backend/utils/misc/guc_tables.c +++ b/src/backend/utils/misc/guc_tables.c @@ -491,6 +491,13 @@ static const struct config_enum_entry file_copy_method_options[] = { {NULL, 0, false} }; +static const struct config_enum_entry numa_options[] = { + {"off", NUMA_OFF, false}, + {"on", NUMA_ON, false}, + {"auto", NUMA_AUTO, false}, + {NULL, 0, false} +}; + /* * Options for enum values stored in other modules */ @@ -579,6 +586,7 @@ static int ssl_renegotiation_limit; int huge_pages = HUGE_PAGES_TRY; int huge_page_size; int huge_pages_status = HUGE_PAGES_UNKNOWN; +int numa = DEFAULT_NUMA; /* * These variables are all dummies that don't do anything, except in some @@ -5418,6 +5426,16 @@ struct config_enum ConfigureNamesEnum[] = NULL, assign_io_method, NULL }, + { + {"numa", PGC_POSTMASTER, RESOURCES_MEM, + gettext_noop("Whether to use NUMA interleaving for shared memory."), + NULL + }, + &numa, + DEFAULT_NUMA, numa_options, + NULL, NULL, NULL + }, + /* End-of-list marker */ { {NULL, 0, 0, NULL, NULL}, NULL, 0, NULL, NULL, NULL, NULL diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 34826d01380..f46b2d8de4d 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -137,6 +137,8 @@ # (change requires restart) #huge_page_size = 0 # zero for system default # (change requires restart) +#numa = off # off, on, auto + # (change requires restart) #temp_buffers = 8MB # min 800kB #max_prepared_transactions = 0 # zero disables the feature # (change requires restart) diff --git a/src/include/port/pg_numa.h b/src/include/port/pg_numa.h index 40f1d324dcf..129663de2e8 100644 --- a/src/include/port/pg_numa.h +++ b/src/include/port/pg_numa.h @@ -17,6 +17,7 @@ extern PGDLLIMPORT int pg_numa_init(void); extern PGDLLIMPORT int pg_numa_query_pages(int pid, unsigned long count, void **pages, int *status); extern PGDLLIMPORT int pg_numa_get_max_node(void); +extern PGDLLIMPORT int pg_numa_interleave_memptr(void *ptr, size_t sz); #ifdef USE_LIBNUMA diff --git a/src/include/storage/pg_shmem.h b/src/include/storage/pg_shmem.h index 5f7d4b83a60..1b09f7ec390 100644 --- a/src/include/storage/pg_shmem.h +++ b/src/include/storage/pg_shmem.h @@ -46,6 +46,7 @@ extern PGDLLIMPORT int shared_memory_type; extern PGDLLIMPORT int huge_pages; extern PGDLLIMPORT int huge_page_size; extern PGDLLIMPORT int huge_pages_status; +extern PGDLLIMPORT int numa; /* Possible values for huge_pages and huge_pages_status */ typedef enum @@ -64,6 +65,15 @@ typedef enum SHMEM_TYPE_MMAP, } PGShmemType; +typedef enum +{ + NUMA_OFF, + NUMA_ON, + NUMA_AUTO, +} NumaType; + +#define DEFAULT_NUMA NUMA_OFF + #ifndef WIN32 extern PGDLLIMPORT unsigned long UsedShmemSegID; #else diff --git a/src/port/pg_numa.c b/src/port/pg_numa.c index 4b487a2a4e8..6ed0a5d2949 100644 --- a/src/port/pg_numa.c +++ b/src/port/pg_numa.c @@ -55,6 +55,17 @@ pg_numa_get_max_node(void) return numa_max_node(); } +int +pg_numa_interleave_memptr(void *ptr, size_t sz) +{ + struct bitmask *nodemask = numa_allocate_nodemask(); + + numa_bitmask_setall(nodemask); + numa_interleave_memory(ptr, sz, nodemask); + numa_free_nodemask(nodemask); + return 0; +} + #else /* Empty wrappers */ @@ -77,4 +88,10 @@ pg_numa_get_max_node(void) return 0; } +int +pg_numa_interleave_memptr(void *ptr, size_t sz) +{ + return 0; +} + #endif -- 2.39.5