Linux-mm Archive on lore.kernel.org
 help / color / Atom feed
From: Mel Gorman <mgorman@suse.de>
To: Peter Zijlstra <a.p.zijlstra@chello.nl>, Rik van Riel <riel@redhat.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>,
	Ingo Molnar <mingo@kernel.org>,
	Andrea Arcangeli <aarcange@redhat.com>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Linux-MM <linux-mm@kvack.org>,
	LKML <linux-kernel@vger.kernel.org>, Mel Gorman <mgorman@suse.de>
Subject: [PATCH 61/63] sched: numa: skip some page migrations after a shared fault
Date: Mon,  7 Oct 2013 11:29:39 +0100
Message-ID: <1381141781-10992-62-git-send-email-mgorman@suse.de> (raw)
In-Reply-To: <1381141781-10992-1-git-send-email-mgorman@suse.de>

From: Rik van Riel <riel@redhat.com>

Shared faults can lead to lots of unnecessary page migrations,
slowing down the system, and causing private faults to hit the
per-pgdat migration ratelimit.

This patch adds sysctl numa_balancing_migrate_deferred, which specifies
how many shared page migrations to skip unconditionally, after each page
migration that is skipped because it is a shared fault.

This reduces the number of page migrations back and forth in
shared fault situations. It also gives a strong preference to
the tasks that are already running where most of the memory is,
and to moving the other tasks to near the memory.

Testing this with a much higher scan rate than the default
still seems to result in fewer page migrations than before.

Memory seems to be somewhat better consolidated than previously,
with multi-instance specjbb runs on a 4 node system.

Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
---
 Documentation/sysctl/kernel.txt | 10 ++++++++-
 include/linux/sched.h           |  5 ++++-
 kernel/sched/fair.c             |  8 +++++++
 kernel/sysctl.c                 |  7 ++++++
 mm/mempolicy.c                  | 48 ++++++++++++++++++++++++++++++++++++++++-
 5 files changed, 75 insertions(+), 3 deletions(-)

diff --git a/Documentation/sysctl/kernel.txt b/Documentation/sysctl/kernel.txt
index 84f1780..4273b2d 100644
--- a/Documentation/sysctl/kernel.txt
+++ b/Documentation/sysctl/kernel.txt
@@ -375,7 +375,8 @@ feature should be disabled. Otherwise, if the system overhead from the
 feature is too high then the rate the kernel samples for NUMA hinting
 faults may be controlled by the numa_balancing_scan_period_min_ms,
 numa_balancing_scan_delay_ms, numa_balancing_scan_period_max_ms,
-numa_balancing_scan_size_mb and numa_balancing_settle_count sysctls.
+numa_balancing_scan_size_mb, numa_balancing_settle_count sysctls and
+numa_balancing_migrate_deferred.
 
 ==============================================================
 
@@ -421,6 +422,13 @@ the schedule balancer stops pushing the task towards a preferred node. This
 gives the scheduler a chance to place the task on an alternative node if the
 preferred node is overloaded.
 
+numa_balancing_migrate_deferred is how many page migrations get skipped
+unconditionally, after a page migration is skipped because a page is shared
+with other tasks. This reduces page migration overhead, and determines
+how much stronger the "move task near its memory" policy scheduler becomes,
+versus the "move memory near its task" memory management policy, for workloads
+with shared memory.
+
 ==============================================================
 
 osrelease, ostype & version:
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 0f6b1b3..b737b72 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1332,6 +1332,8 @@ struct task_struct {
 	int numa_scan_seq;
 	unsigned int numa_scan_period;
 	unsigned int numa_scan_period_max;
+	int numa_preferred_nid;
+	int numa_migrate_deferred;
 	unsigned long numa_migrate_retry;
 	u64 node_stamp;			/* migration stamp  */
 	struct callback_head numa_work;
@@ -1362,7 +1364,6 @@ struct task_struct {
 	 */
 	unsigned long numa_faults_locality[2];
 
-	int numa_preferred_nid;
 	unsigned long numa_pages_migrated;
 #endif /* CONFIG_NUMA_BALANCING */
 
@@ -1459,6 +1460,8 @@ extern void task_numa_fault(int last_node, int node, int pages, int flags);
 extern pid_t task_numa_group_id(struct task_struct *p);
 extern void set_numabalancing_state(bool enabled);
 extern void task_numa_free(struct task_struct *p);
+
+extern unsigned int sysctl_numa_balancing_migrate_deferred;
 #else
 static inline void task_numa_fault(int last_node, int node, int pages,
 				   int flags)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index ceffce9..9e2271b 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -833,6 +833,14 @@ unsigned int sysctl_numa_balancing_scan_size = 256;
 /* Scan @scan_size MB every @scan_period after an initial @scan_delay in ms */
 unsigned int sysctl_numa_balancing_scan_delay = 1000;
 
+/*
+ * After skipping a page migration on a shared page, skip N more numa page
+ * migrations unconditionally. This reduces the number of NUMA migrations
+ * in shared memory workloads, and has the effect of pulling tasks towards
+ * where their memory lives, over pulling the memory towards the task.
+ */
+unsigned int sysctl_numa_balancing_migrate_deferred = 16;
+
 static unsigned int task_nr_scan_windows(struct task_struct *p)
 {
 	unsigned long rss = 0;
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index e509b90..a159e1f 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -391,6 +391,13 @@ static struct ctl_table kern_table[] = {
 		.mode           = 0644,
 		.proc_handler   = proc_dointvec,
 	},
+	{
+		.procname       = "numa_balancing_migrate_deferred",
+		.data           = &sysctl_numa_balancing_migrate_deferred,
+		.maxlen         = sizeof(unsigned int),
+		.mode           = 0644,
+		.proc_handler   = proc_dointvec,
+	},
 #endif /* CONFIG_NUMA_BALANCING */
 #endif /* CONFIG_SCHED_DEBUG */
 	{
diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 2929c24..71cb253 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -2301,6 +2301,35 @@ static void sp_free(struct sp_node *n)
 	kmem_cache_free(sn_cache, n);
 }
 
+#ifdef CONFIG_NUMA_BALANCING
+static bool numa_migrate_deferred(struct task_struct *p, int last_cpupid)
+{
+	/* Never defer a private fault */
+	if (cpupid_match_pid(p, last_cpupid))
+		return false;
+
+	if (p->numa_migrate_deferred) {
+		p->numa_migrate_deferred--;
+		return true;
+	}
+	return false;
+}
+
+static inline void defer_numa_migrate(struct task_struct *p)
+{
+	p->numa_migrate_deferred = sysctl_numa_balancing_migrate_deferred;
+}
+#else
+static inline bool numa_migrate_deferred(struct task_struct *p, int last_cpupid)
+{
+	return false;
+}
+
+static inline void defer_numa_migrate(struct task_struct *p)
+{
+}
+#endif /* CONFIG_NUMA_BALANCING */
+
 /**
  * mpol_misplaced - check whether current page node is valid in policy
  *
@@ -2402,7 +2431,24 @@ int mpol_misplaced(struct page *page, struct vm_area_struct *vma, unsigned long
 		 * relation.
 		 */
 		last_cpupid = page_cpupid_xchg_last(page, this_cpupid);
-		if (!cpupid_pid_unset(last_cpupid) && cpupid_to_nid(last_cpupid) != thisnid)
+		if (!cpupid_pid_unset(last_cpupid) && cpupid_to_nid(last_cpupid) != thisnid) {
+
+			/* See sysctl_numa_balancing_migrate_deferred comment */
+			if (!cpupid_match_pid(current, last_cpupid))
+				defer_numa_migrate(current);
+
+			goto out;
+		}
+
+		/*
+		 * The quadratic filter above reduces extraneous migration
+		 * of shared pages somewhat. This code reduces it even more,
+		 * reducing the overhead of page migrations of shared pages.
+		 * This makes workloads with shared pages rely more on
+		 * "move task near its memory", and less on "move memory
+		 * towards its task", which is exactly what we want.
+		 */
+		if (numa_migrate_deferred(current, last_cpupid))
 			goto out;
 	}
 
-- 
1.8.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  parent reply index

Thread overview: 136+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-10-07 10:28 [PATCH 0/63] Basic scheduler support for automatic NUMA balancing V9 Mel Gorman
2013-10-07 10:28 ` [PATCH 01/63] hotplug: Optimize {get,put}_online_cpus() Mel Gorman
2013-10-07 10:28 ` [PATCH 02/63] mm: numa: Document automatic NUMA balancing sysctls Mel Gorman
2013-10-07 12:46   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 03/63] sched, numa: Comment fixlets Mel Gorman
2013-10-07 12:46   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 04/63] mm: numa: Do not account for a hinting fault if we raced Mel Gorman
2013-10-07 12:47   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 05/63] mm: Wait for THP migrations to complete during NUMA hinting faults Mel Gorman
2013-10-07 13:55   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 06/63] mm: Prevent parallel splits during THP migration Mel Gorman
2013-10-07 14:01   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 07/63] mm: numa: Sanitize task_numa_fault() callsites Mel Gorman
2013-10-07 14:02   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 08/63] mm: Close races between THP migration and PMD numa clearing Mel Gorman
2013-10-07 14:02   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 09/63] mm: Account for a THP NUMA hinting update as one PTE update Mel Gorman
2013-10-07 14:02   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 10/63] mm: Do not flush TLB during protection change if !pte_present && !migration_entry Mel Gorman
2013-10-07 15:12   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 11/63] mm: Only flush TLBs if a transhuge PMD is modified for NUMA pte scanning Mel Gorman
2013-10-07 10:28 ` [PATCH 12/63] mm: numa: Do not migrate or account for hinting faults on the zero page Mel Gorman
2013-10-07 17:10   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 13/63] sched: numa: Mitigate chance that same task always updates PTEs Mel Gorman
2013-10-07 17:24   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 14/63] sched: numa: Continue PTE scanning even if migrate rate limited Mel Gorman
2013-10-07 17:24   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 15/63] Revert "mm: sched: numa: Delay PTE scanning until a task is scheduled on a new node" Mel Gorman
2013-10-07 17:42   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 16/63] sched: numa: Initialise numa_next_scan properly Mel Gorman
2013-10-07 17:44   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 17/63] sched: Set the scan rate proportional to the memory usage of the task being scanned Mel Gorman
2013-10-07 17:44   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 18/63] sched: numa: Slow scan rate if no NUMA hinting faults are being recorded Mel Gorman
2013-10-07 18:02   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 19/63] sched: Track NUMA hinting faults on per-node basis Mel Gorman
2013-10-07 18:02   ` Rik van Riel
2013-12-04  5:32   ` Wanpeng Li
2013-12-04  5:37     ` Wanpeng Li
2013-10-07 10:28 ` [PATCH 20/63] sched: Select a preferred node with the most numa hinting faults Mel Gorman
2013-10-07 18:04   ` Rik van Riel
2013-10-07 10:28 ` [PATCH 21/63] sched: Update NUMA hinting faults once per scan Mel Gorman
2013-10-07 18:39   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 22/63] sched: Favour moving tasks towards the preferred node Mel Gorman
2013-10-07 18:39   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 23/63] sched: Resist moving tasks towards nodes with fewer hinting faults Mel Gorman
2013-10-07 18:40   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 24/63] sched: Reschedule task on preferred NUMA node once selected Mel Gorman
2013-10-07 18:40   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 25/63] sched: Add infrastructure for split shared/private accounting of NUMA hinting faults Mel Gorman
2013-10-07 18:41   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 26/63] sched: Check current->mm before allocating NUMA faults Mel Gorman
2013-10-07 18:41   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 27/63] mm: numa: Scan pages with elevated page_mapcount Mel Gorman
2013-10-07 18:43   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 28/63] sched: Remove check that skips small VMAs Mel Gorman
2013-10-07 18:44   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 29/63] sched: Set preferred NUMA node based on number of private faults Mel Gorman
2013-10-07 18:45   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 30/63] sched: Do not migrate memory immediately after switching node Mel Gorman
2013-10-07 10:29 ` [PATCH 31/63] mm: numa: only unmap migrate-on-fault VMAs Mel Gorman
2013-10-07 10:29 ` [PATCH 32/63] sched: Avoid overloading CPUs on a preferred NUMA node Mel Gorman
2013-10-07 18:58   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 33/63] sched: Retry migration of tasks to CPU on a preferred node Mel Gorman
2013-10-07 18:58   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 34/63] sched: numa: increment numa_migrate_seq when task runs in correct location Mel Gorman
2013-10-07 10:29 ` [PATCH 35/63] sched: numa: Do not trap hinting faults for shared libraries Mel Gorman
2013-10-07 19:04   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 36/63] mm: numa: Only trap pmd hinting faults if we would otherwise trap PTE faults Mel Gorman
2013-10-07 19:06   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 37/63] stop_machine: Introduce stop_two_cpus() Mel Gorman
2013-10-07 10:29 ` [PATCH 38/63] sched: Introduce migrate_swap() Mel Gorman
2013-10-07 19:06   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 39/63] sched: numa: Use a system-wide search to find swap/migration candidates Mel Gorman
2013-10-07 19:07   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 40/63] sched: numa: Favor placing a task on the preferred node Mel Gorman
2013-10-07 19:07   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 41/63] sched: numa: fix placement of workloads spread across multiple nodes Mel Gorman
2013-10-07 10:29 ` [PATCH 42/63] mm: numa: Change page last {nid,pid} into {cpu,pid} Mel Gorman
2013-10-07 19:08   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 43/63] sched: numa: Use {cpu, pid} to create task groups for shared faults Mel Gorman
2013-10-07 19:09   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 44/63] sched: numa: Report a NUMA task group ID Mel Gorman
2013-10-07 19:09   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 45/63] mm: numa: copy cpupid on page migration Mel Gorman
2013-10-07 10:29 ` [PATCH 46/63] mm: numa: Do not group on RO pages Mel Gorman
2013-10-07 19:10   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 47/63] mm: numa: Do not batch handle PMD pages Mel Gorman
2013-10-07 19:11   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 48/63] sched: numa: stay on the same node if CLONE_VM Mel Gorman
2013-10-07 10:29 ` [PATCH 49/63] sched: numa: use group fault statistics in numa placement Mel Gorman
2013-10-07 10:29 ` [PATCH 50/63] sched: numa: call task_numa_free from do_execve Mel Gorman
2013-10-07 10:29 ` [PATCH 51/63] sched: numa: Prevent parallel updates to group stats during placement Mel Gorman
2013-10-07 19:13   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 52/63] sched: numa: add debugging Mel Gorman
2013-10-07 19:13   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 53/63] sched: numa: Decide whether to favour task or group weights based on swap candidate relationships Mel Gorman
2013-10-07 10:29 ` [PATCH 54/63] sched: numa: fix task or group comparison Mel Gorman
2013-10-07 10:29 ` [PATCH 55/63] sched: numa: Avoid migrating tasks that are placed on their preferred node Mel Gorman
2013-10-07 19:14   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 56/63] sched: numa: be more careful about joining numa groups Mel Gorman
2013-10-07 10:29 ` [PATCH 57/63] sched: numa: Take false sharing into account when adapting scan rate Mel Gorman
2013-10-07 19:14   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 58/63] sched: numa: adjust scan rate in task_numa_placement Mel Gorman
2013-10-07 10:29 ` [PATCH 59/63] sched: numa: Remove the numa_balancing_scan_period_reset sysctl Mel Gorman
2013-10-07 19:14   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 60/63] mm: numa: revert temporarily disabling of NUMA migration Mel Gorman
2013-10-07 10:29 ` Mel Gorman [this message]
2013-10-07 10:29 ` [PATCH 62/63] sched: numa: use unsigned longs for numa group fault stats Mel Gorman
2013-10-07 19:15   ` Rik van Riel
2013-10-07 10:29 ` [PATCH 63/63] sched: numa: periodically retry task_numa_migrate Mel Gorman
2013-10-09 11:03 ` [PATCH 0/63] Basic scheduler support for automatic NUMA balancing V9 Ingo Molnar
2013-10-09 11:11   ` Ingo Molnar
2013-10-09 11:13     ` Ingo Molnar
2013-10-09 12:05   ` Peter Zijlstra
2013-10-09 12:48     ` Ingo Molnar
2013-10-10  7:05   ` Mel Gorman
2013-10-09 16:28 ` Ingo Molnar
2013-10-09 16:29   ` Ingo Molnar
2013-10-09 16:57     ` Ingo Molnar
2013-10-09 17:09       ` Ingo Molnar
2013-10-09 17:11         ` Peter Zijlstra
2013-10-09 17:08   ` Peter Zijlstra
2013-10-09 17:15     ` Ingo Molnar
2013-10-09 17:18       ` Peter Zijlstra
2013-10-24 12:26 ` Automatic NUMA balancing patches for tip-urgent/stable Mel Gorman
2013-10-26 12:11   ` Ingo Molnar
2013-10-29  9:42     ` Mel Gorman
2013-10-29  9:48       ` Ingo Molnar
2013-10-29 10:24         ` Mel Gorman
2013-10-29 10:41           ` Ingo Molnar
2013-10-29 12:48             ` Mel Gorman
2013-10-31  9:51   ` [RFC GIT PULL] NUMA-balancing memory corruption fixes Ingo Molnar
2013-10-31 22:25     ` Linus Torvalds
2013-11-01  7:36       ` Ingo Molnar
  -- strict thread matches above, loose matches on Subject: below --
2013-09-27 13:26 [PATCH 0/63] Basic scheduler support for automatic NUMA balancing V8 Mel Gorman
2013-09-27 13:27 ` [PATCH 61/63] sched: numa: skip some page migrations after a shared fault Mel Gorman

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1381141781-10992-62-git-send-email-mgorman@suse.de \
    --to=mgorman@suse.de \
    --cc=a.p.zijlstra@chello.nl \
    --cc=aarcange@redhat.com \
    --cc=hannes@cmpxchg.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mingo@kernel.org \
    --cc=riel@redhat.com \
    --cc=srikar@linux.vnet.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-mm Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-mm/0 linux-mm/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-mm linux-mm/ https://lore.kernel.org/linux-mm \
		linux-mm@kvack.org
	public-inbox-index linux-mm

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kvack.linux-mm


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git