btrfs: introduce size class to block group allocator

The aim of this patch is to reduce the fragmentation of block groups
under certain unhappy workloads. It is particularly effective when the
size of extents correlates with their lifetime, which is something we
have observed causing fragmentation in the fleet at Meta.

This patch categorizes extents into size classes:

- x < 128KiB: "small"
- 128KiB < x < 8MiB: "medium"
- x > 8MiB: "large"

and as much as possible reduces allocations of extents into block groups
that don't match the size class. This takes advantage of any (possible)
correlation between size and lifetime and also leaves behind predictable
re-usable gaps when extents are freed; small writes don't gum up bigger
holes.

Size classes are implemented in the following way:

- Mark each new block group with a size class of the first allocation
  that goes into it.

- Add two new passes to ffe: "unset size class" and "wrong size class".
  First, try only matching block groups, then try unset ones, then allow
  allocation of new ones, and finally allow mismatched block groups.

- Filtering is done just by skipping inappropriate ones, there is no
  special size class indexing.

Other solutions I considered were:

- A best fit allocator with an rb-tree. This worked well, as small
  writes didn't leak big holes from large freed extents, but led to
  regressions in ffe and write performance due to lock contention on
  the rb-tree with every allocation possibly updating it in parallel.
  Perhaps something clever could be done to do the updates in the
  background while being "right enough".

- A fixed size "working set". This prevents freeing an extent
  drastically changing where writes currently land, and seems like a
  good option too. Doesn't take advantage of size in any way.

- The same size class idea, but implemented with xarray marks. This
  turned out to be slower than looping the linked list and skipping
  wrong block groups, and is also less flexible since we must have only
  3 size classes (max #marks). With the current approach we can have as
  many as we like.

Performance testing was done via: https://github.com/josefbacik/fsperf
Of particular relevance are the new fragmentation specific tests.

A brief summary of the testing results:

- Neutral results on existing tests. There are some minor regressions
  and improvements here and there, but nothing that truly stands out as
  notable.
- Improvement on new tests where size class and extent lifetime are
  correlated. Fragmentation in these cases is completely eliminated
  and write performance is generally a little better. There is also
  significant improvement where extent sizes are just a bit larger than
  the size class boundaries.
- Regression on one new tests: where the allocations are sized
  intentionally a hair under the borders of the size classes. Results
  are neutral on the test that intentionally attacks this new scheme by
  mixing extent size and lifetime.

The full dump of the performance results can be found here:
https://bur.io/fsperf/size-class-2022-11-15.txt
(there are ANSI escape codes, so best to curl and view in terminal)

Here is a snippet from the full results for a new test which mixes
buffered writes appending to a long lived set of files and large short
lived fallocates:

bufferedappendvsfallocate results
         metric             baseline       current        stdev            diff
======================================================================================
avg_commit_ms                    31.13         29.20          2.67     -6.22%
bg_count                            14         15.60             0     11.43%
commits                          11.10         12.20          0.32      9.91%
elapsed                          27.30         26.40          2.98     -3.30%
end_state_mount_ns         11122551.90   10635118.90     851143.04     -4.38%
end_state_umount_ns           1.36e+09      1.35e+09   12248056.65     -1.07%
find_free_extent_calls       116244.30     114354.30        964.56     -1.63%
find_free_extent_ns_max      599507.20    1047168.20     103337.08     74.67%
find_free_extent_ns_mean       3607.19       3672.11        101.20      1.80%
find_free_extent_ns_min            500           512          6.67      2.40%
find_free_extent_ns_p50           2848          2876         37.65      0.98%
find_free_extent_ns_p95           4916          5000         75.45      1.71%
find_free_extent_ns_p99       20734.49      20920.48       1670.93      0.90%
frag_pct_max                     61.67             0          8.05   -100.00%
frag_pct_mean                    43.59             0          6.10   -100.00%
frag_pct_min                     25.91             0         16.60   -100.00%
frag_pct_p50                     42.53             0          7.25   -100.00%
frag_pct_p95                     61.67             0          8.05   -100.00%
frag_pct_p99                     61.67             0          8.05   -100.00%
fragmented_bg_count               6.10             0          1.45   -100.00%
max_commit_ms                    49.80            46          5.37     -7.63%
sys_cpu                           2.59          2.62          0.29      1.39%
write_bw_bytes                1.62e+08      1.68e+08   17975843.50      3.23%
write_clat_ns_mean            57426.39      54475.95       2292.72     -5.14%
write_clat_ns_p50             46950.40      42905.60       2101.35     -8.62%
write_clat_ns_p99            148070.40     143769.60       2115.17     -2.90%
write_io_kbytes                4194304       4194304             0      0.00%
write_iops                     2476.15       2556.10        274.29      3.23%
write_lat_ns_max            2101667.60    2251129.50     370556.59      7.11%
write_lat_ns_mean             59374.91      55682.00       2523.09     -6.22%
write_lat_ns_min              17353.10         16250       1646.08     -6.36%

There are some mixed improvements/regressions in most metrics along with
an elimination of fragmentation in this workload.

On the balance, the drastic 1->0 improvement in the happy cases seems
worth the mix of regressions and improvements we do observe.

Some considerations for future work:

- Experimenting with more size classes
- More hinting/search ordering work to approximate a best-fit allocator

Signed-off-by: Boris Burkov <boris@bur.io>
Signed-off-by: David Sterba <dsterba@suse.com>

1 files changed, 89 insertions, 16 deletions

diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c
index e90800388a41..6557b1b7f89a 100644
--- a/fs/btrfs/block-group.c
+++ b/fs/btrfs/block-group.c
@@ -1,5 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0
 
+#include <linux/sizes.h>
 #include <linux/list_sort.h>
 #include "misc.h"
 #include "ctree.h"
@@ -3379,6 +3380,7 @@ int btrfs_update_block_group(struct btrfs_trans_handle *trans,
 			cache->space_info->disk_used -= num_bytes * factor;
 
 			reclaim = should_reclaim_block_group(cache, num_bytes);
+
 			spin_unlock(&cache->lock);
 			spin_unlock(&cache->space_info->lock);
 
@@ -3433,32 +3435,42 @@ int btrfs_update_block_group(struct btrfs_trans_handle *trans,
  * reservation and return -EAGAIN, otherwise this function always succeeds.
  */
 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
-			     u64 ram_bytes, u64 num_bytes, int delalloc)
+			     u64 ram_bytes, u64 num_bytes, int delalloc,
+			     bool force_wrong_size_class)
 {
 	struct btrfs_space_info *space_info = cache->space_info;
+	enum btrfs_block_group_size_class size_class;
 	int ret = 0;
 
 	spin_lock(&space_info->lock);
 	spin_lock(&cache->lock);
 	if (cache->ro) {
 		ret = -EAGAIN;
-	} else {
-		cache->reserved += num_bytes;
-		space_info->bytes_reserved += num_bytes;
-		trace_btrfs_space_reservation(cache->fs_info, "space_info",
-					      space_info->flags, num_bytes, 1);
-		btrfs_space_info_update_bytes_may_use(cache->fs_info,
-						      space_info, -ram_bytes);
-		if (delalloc)
-			cache->delalloc_bytes += num_bytes;
+		goto out;
+	}
 
-		/*
-		 * Compression can use less space than we reserved, so wake
-		 * tickets if that happens
-		 */
-		if (num_bytes < ram_bytes)
-			btrfs_try_granting_tickets(cache->fs_info, space_info);
+	if (btrfs_is_block_group_data_only(cache)) {
+		size_class = btrfs_calc_block_group_size_class(num_bytes);
+		ret = btrfs_use_block_group_size_class(cache, size_class, force_wrong_size_class);
+		if (ret)
+			goto out;
 	}
+	cache->reserved += num_bytes;
+	space_info->bytes_reserved += num_bytes;
+	trace_btrfs_space_reservation(cache->fs_info, "space_info",
+				      space_info->flags, num_bytes, 1);
+	btrfs_space_info_update_bytes_may_use(cache->fs_info,
+					      space_info, -ram_bytes);
+	if (delalloc)
+		cache->delalloc_bytes += num_bytes;
+
+	/*
+	 * Compression can use less space than we reserved, so wake tickets if
+	 * that happens.
+	 */
+	if (num_bytes < ram_bytes)
+		btrfs_try_granting_tickets(cache->fs_info, space_info);
+out:
 	spin_unlock(&cache->lock);
 	spin_unlock(&space_info->lock);
 	return ret;
@@ -4218,3 +4230,64 @@ void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount
 	bg->swap_extents -= amount;
 	spin_unlock(&bg->lock);
 }
+
+enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size)
+{
+	if (size <= SZ_128K)
+		return BTRFS_BG_SZ_SMALL;
+	if (size <= SZ_8M)
+		return BTRFS_BG_SZ_MEDIUM;
+	return BTRFS_BG_SZ_LARGE;
+}
+
+/*
+ * Handle a block group allocating an extent in a size class
+ *
+ * @bg:				The block group we allocated in.
+ * @size_class:			The size class of the allocation.
+ * @force_wrong_size_class:	Whether we are desperate enough to allow
+ *				mismatched size classes.
+ *
+ * Returns: 0 if the size class was valid for this block_group, -EAGAIN in the
+ * case of a race that leads to the wrong size class without
+ * force_wrong_size_class set.
+ *
+ * find_free_extent will skip block groups with a mismatched size class until
+ * it really needs to avoid ENOSPC. In that case it will set
+ * force_wrong_size_class. However, if a block group is newly allocated and
+ * doesn't yet have a size class, then it is possible for two allocations of
+ * different sizes to race and both try to use it. The loser is caught here and
+ * has to retry.
+ */
+int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
+				     enum btrfs_block_group_size_class size_class,
+				     bool force_wrong_size_class)
+{
+	ASSERT(size_class != BTRFS_BG_SZ_NONE);
+
+	/* The new allocation is in the right size class, do nothing */
+	if (bg->size_class == size_class)
+		return 0;
+	/*
+	 * The new allocation is in a mismatched size class.
+	 * This means one of two things:
+	 *
+	 * 1. Two tasks in find_free_extent for different size_classes raced
+	 *    and hit the same empty block_group. Make the loser try again.
+	 * 2. A call to find_free_extent got desperate enough to set
+	 *    'force_wrong_slab'. Don't change the size_class, but allow the
+	 *    allocation.
+	 */
+	if (bg->size_class != BTRFS_BG_SZ_NONE) {
+		if (force_wrong_size_class)
+			return 0;
+		return -EAGAIN;
+	}
+	/*
+	 * The happy new block group case: the new allocation is the first
+	 * one in the block_group so we set size_class.
+	 */
+	bg->size_class = size_class;
+
+	return 0;
+}