Index: linux-2.6.16-rc1-mm2/include/linux/swap.h
===================================================================
--- linux-2.6.16-rc1-mm2.orig/include/linux/swap.h	2006-01-20 08:35:10.000000000 -0800
+++ linux-2.6.16-rc1-mm2/include/linux/swap.h	2006-01-23 10:08:29.000000000 -0800
@@ -178,6 +178,7 @@ extern int vm_swappiness;
 
 #ifdef CONFIG_NUMA
 extern int zone_reclaim_mode;
+extern int zone_reclaim_interval;
 extern int zone_reclaim(struct zone *, gfp_t, unsigned int);
 #else
 #define zone_reclaim_mode 0
Index: linux-2.6.16-rc1-mm2/mm/vmscan.c
===================================================================
--- linux-2.6.16-rc1-mm2.orig/mm/vmscan.c	2006-01-23 10:02:23.000000000 -0800
+++ linux-2.6.16-rc1-mm2/mm/vmscan.c	2006-01-23 10:09:27.000000000 -0800
@@ -1827,10 +1827,15 @@ module_init(kswapd_init)
  */
 int zone_reclaim_mode __read_mostly;
 
+#define RECLAIM_OFF 0
+#define RECLAIM_ZONE (1<<0)	/* Run shrink_cache on the zone */
+#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
+#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */
+
 /*
  * Mininum time between zone reclaim scans
  */
-#define ZONE_RECLAIM_INTERVAL 30*HZ
+int zone_reclaim_interval __read_mostly = 30*HZ;
 
 /*
  * Priority for ZONE_RECLAIM. This determines the fraction of pages
@@ -1850,7 +1855,7 @@ int zone_reclaim(struct zone *zone, gfp_
 	struct scan_control sc;
 
 	if (time_before(jiffies,
-		zone->last_unsuccessful_zone_reclaim + ZONE_RECLAIM_INTERVAL))
+		zone->last_unsuccessful_zone_reclaim + zone_reclaim_interval))
 			return 0;
 
 	if (!(gfp_mask & __GFP_WAIT) ||
@@ -1860,8 +1865,8 @@ int zone_reclaim(struct zone *zone, gfp_
 		atomic_read(&zone->reclaim_in_progress) > 0)
 			return 0;
 
-	sc.may_writepage = 0;
-	sc.may_swap = 0;
+	sc.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE);
+	sc.may_swap = !!(zone_reclaim_mode & RECLAIM_SWAP);
 	sc.nr_scanned = 0;
 	sc.nr_reclaimed = 0;
 	sc.priority = ZONE_RECLAIM_PRIORITY + 1;
Index: linux-2.6.16-rc1-mm2/Documentation/sysctl/vm.txt
===================================================================
--- linux-2.6.16-rc1-mm2.orig/Documentation/sysctl/vm.txt	2006-01-20 08:35:07.000000000 -0800
+++ linux-2.6.16-rc1-mm2/Documentation/sysctl/vm.txt	2006-01-23 10:08:29.000000000 -0800
@@ -28,6 +28,7 @@ Currently, these files are in /proc/sys/
 - block_dump
 - drop-caches
 - zone_reclaim_mode
+- zone_reclaim_interval
 
 ==============================================================
 
@@ -126,8 +127,18 @@ the high water marks for each per cpu pa
 
 zone_reclaim_mode:
 
-This is set during bootup to 1 if it is determined that pages from
-remote zones will cause a significant performance reduction. The
+This allows to set more or less agressive forms of reclaiming memory
+when a zone runs out of memory. If it is set to zero then no
+zone reclaim occurs.
+
+This is a ORed together value of
+
+1	= Zone reclaim on
+2	= Zone reclaim writes dirty pages out
+4	= Zone reclaim swaps pages
+
+zone_reclaim_mode is set during bootup to 1 if it is determined that pages
+from remote zones will cause a significant performance reduction. The
 page allocator will then reclaim easily reusable pages (those page
 cache pages that are currently not used) before going off node.
 
@@ -135,6 +146,27 @@ The user can override this setting. It m
 off zone reclaim if the system is used for a file server and all
 of memory should be used for caching files from disk.
 
-It may be beneficial to switch this on if one wants to do zone
-reclaim regardless of the numa distances in the system.
+Allowing zone reclaim to write out pages stops an process doing huge
+amounts of writes from dirtying pages on other nodes. Instead the
+process will be throttled when zone reclaim begins to writeout
+the locally dirtied pages. This may decrease the performance
+of a single process since it cannot use all of system memory to buffer
+the outgoing writes anymore but it will protect processes on other
+nodes from the writeout happy process.
+
+Allowing regular swap effectively restricts allocations to the local
+node unless explicitly overridden by memory policy.
+
+================================================================
+
+zone_reclaim_interval:
+
+The time allowed for off node allocations after zone reclaim
+has failed to reclaim enough pages to allow a local allocation.
+
+This is 30 seconds by default. Time is set in ticks.
+F.e. if HZ = 250 then 30 seconds is 30*250 = 7500.
+
+If undesired off node allocations occur then this interval should be
+reduced. However, too frequent scans will degrade performance.
 
Index: linux-2.6.16-rc1-mm2/kernel/sysctl.c
===================================================================
--- linux-2.6.16-rc1-mm2.orig/kernel/sysctl.c	2006-01-20 16:30:06.000000000 -0800
+++ linux-2.6.16-rc1-mm2/kernel/sysctl.c	2006-01-23 10:08:29.000000000 -0800
@@ -889,6 +889,16 @@ static ctl_table vm_table[] = {
 		.strategy	= &sysctl_intvec,
 		.extra1		= &zero,
 	},
+	{
+		.ctl_name	= VM_ZONE_RECLAIM_INTERVAL,
+		.procname	= "zone_reclaim_interval",
+		.data		= &zone_reclaim_interval,
+		.maxlen		= sizeof(zone_reclaim_interval),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+		.strategy	= &sysctl_intvec,
+		.extra1		= &one_hundred,
+	},
 #endif
 	{ .ctl_name = 0 }
 };
Index: linux-2.6.16-rc1-mm2/include/linux/sysctl.h
===================================================================
--- linux-2.6.16-rc1-mm2.orig/include/linux/sysctl.h	2006-01-20 08:35:10.000000000 -0800
+++ linux-2.6.16-rc1-mm2/include/linux/sysctl.h	2006-01-23 10:08:29.000000000 -0800
@@ -184,6 +184,7 @@ enum
 	VM_DROP_PAGECACHE=29,	/* int: nuke lots of pagecache */
 	VM_PERCPU_PAGELIST_FRACTION=30,/* int: fraction of pages in each percpu_pagelist */
 	VM_ZONE_RECLAIM_MODE=31,/* reclaim local zone memory before going off node */
+	VM_ZONE_RECLAIM_INTERVAL=32,/* time period to wait after reclaim failure */
 };