numa balancer is a module which will try to automatically adjust numa
balancing stuff to gain numa bonus as much as possible.
For each memory cgroup, we process the work in two steps:
On stage 1 we check cgroup's exectime and memory topology to see
if there could be a candidate for settled down, if we got one then
move onto stage 2.
On stage 2 we try to settle down as much as possible by prefer the
candidate node, if the node no longer suitable or locality keep
downturn, we reset things and new round begin.
Decision made with find_candidate_nid(), should_prefer() and keep_prefer(),
which try to pick a candidate node, see if allowed to prefer it and if
keep doing the prefer.
Tested on the box with 96 cpus with sysbench-mysql-oltp_read_write
testing, 4 mysqld instances created and attached to 4 cgroups, 4
sysbench instances then created and attached to corresponding cgroup
to test the mysql with oltp_read_write script, average eps show:
origin balancer
4 instances each 12 threads 5241.08 5375.59 +2.50%
4 instances each 24 threads 7497.29 7820.73 +4.13%
4 instances each 36 threads 8985.44 9317.04 +3.55%
4 instances each 48 threads 9716.50 9982.60 +2.66%
Other benchmark liks dbench, pgbench, perf bench numa also tested, and
with different parameters and number of instances/threads, most of
the cases show bonus, some show acceptable regression, and some got no
changes.
TODO:
* improve the logical to address the regression cases
* Find a way, maybe, to handle the page cache left on remote
* find more scenery which could gain benefit
Signed-off-by: Michael Wang
---
drivers/Makefile | 1 +
drivers/numa/Makefile| 1 +
drivers/numa/numa_balancer.c | 715 +++
3 files changed, 717 insertions(+)
create mode 100644 drivers/numa/Makefile
create mode 100644 drivers/numa/numa_balancer.c
diff --git a/drivers/Makefile b/drivers/Makefile
index c61cde554340..f07936b03870 100644
--- a/drivers/Makefile
+++ b/drivers/Makefile
@@ -187,3 +187,4 @@ obj-$(CONFIG_UNISYS_VISORBUS) += visorbus/
obj-$(CONFIG_SIOX) += siox/
obj-$(CONFIG_GNSS) += gnss/
obj-$(CONFIG_INTERCONNECT) += interconnect/
+obj-$(CONFIG_NUMA_BALANCING) += numa/
diff --git a/drivers/numa/Makefile b/drivers/numa/Makefile
new file mode 100644
index ..acf8a408
--- /dev/null
+++ b/drivers/numa/Makefile
@@ -0,0 +1 @@
+obj-m += numa_balancer.o
diff --git a/drivers/numa/numa_balancer.c b/drivers/numa/numa_balancer.c
new file mode 100644
index ..25bbe08c82a2
--- /dev/null
+++ b/drivers/numa/numa_balancer.c
@@ -0,0 +1,715 @@
+/*
+ * NUMA Balancer
+ *
+ * Copyright (C) 2019 Alibaba Group Holding Limited.
+ * Author: Michael Wang
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+
+static unsigned int debug_level;
+module_param(debug_level, uint, 0644);
+MODULE_PARM_DESC(debug_level, "1 to print decisions, 2 to print both decisions
and node info");
+
+static int prefer_level = 10;
+module_param(prefer_level, int, 0644);
+MODULE_PARM_DESC(prefer_level, "stop numa prefer when reach this much
continuous downturn, 0 means no prefer");
+
+static unsigned int locality_level = PERCENT_70_79;
+module_param(locality_level, uint, 0644);
+MODULE_PARM_DESC(locality_level, "consider locality as good when above this
sector");
+
+static unsigned long period_max = (600 * HZ);
+module_param(period_max, ulong, 0644);
+MODULE_PARM_DESC(period_max, "maximum period between each stage");
+
+static unsigned long period_min = (5 * HZ);
+module_param(period_min, ulong, 0644);
+MODULE_PARM_DESC(period_min, "minimum period between each stage");
+
+static unsigned int cpu_high_wmark = 100;
+module_param(cpu_high_wmark, uint, 0644);
+MODULE_PARM_DESC(cpu_high_wmark, "respect the execution percent rather than
memory percent when above this cpu usage");
+
+static unsigned int cpu_low_wmark = 10;
+module_param(cpu_low_wmark, uint, 0644);
+MODULE_PARM_DESC(cpu_low_wmark, "consider cgroup as active when above this cpu
usage");
+
+static unsigned int free_low_wmark = 10;
+module_param(free_low_wmark, uint, 0644);
+MODULE_PARM_DESC(free_low_wmark, "consider node as consumed out when below
this free percent");
+
+static unsigned int candidate_wmark = 60;
+module_param(candidate_wmark, uint, 0644);
+MODULE_PARM_DESC(candidate_wmark, "consider node as candidate when above this
execution time or memory percent");
+
+static unsigned int settled_wmark = 90;
+module_param(settled_wmark, uint, 0644);
+MODULE_PARM_DESC(settled_wmark, "consider cgroup settle down on node when
above this execution time and memory percent, or