Give the refcount cache the minimum possible size by default

Hi,

we talked about this the other day, so here are the patches to change the default cache sizes in qcow2.

Without this patch:

* refcount-cache-size = l2-cache-size / 4

unless otherwise specified by the user. This is wasteful, the refcount cache is accessed sequentially during normal I/O, so there's no point in caching more tables. I measured the effect on the refcount cache size when populating an empty qcow2 image using random writes, and there's no difference between having the minimum or the maximum sizes(*).

With this patch:

* refcount-cache-size is always 4 clusters by default (the minimum)

* If "cache-size" is set then l2-cache-size is set to the maximum if possible (disk_size * 8 / cluster_size) and the remainder is assigned to the refcount cache.

Regards,

Berto

(*) there is, actually: having a very large cache can even make the
I/O slightly slower, because the larger the cache the longer it takes longer to find a cached entry. I only noticed this under tmpfs anyway.

Changes:
v3:
- Mention that if you use internal snapshots you may want to increase the cache size [Max]

v2: https://lists.gnu.org/archive/html/qemu-block/2018-03/msg00822.html
- s/overriden/overridden/ (in both patches)

v1: https://lists.gnu.org/archive/html/qemu-block/2018-03/msg00709.html
- Initial release

Alberto Garcia (2):
qcow2: Give the refcount cache the minimum possible size by default
docs: Document the new default sizes of the qcow2 caches

 block/qcow2.c              | 31 +++++++++++++++++++------------
 block/qcow2.h              |  4 ----
 docs/qcow2-cache.txt       | 33 ++++++++++++++++-----------------
 tests/qemu-iotests/137.out |  2 +-
 4 files changed, 36 insertions(+), 34 deletions(-)

2.11.0

We have just reduced the refcount cache size to the minimum unless the user explicitly requests a larger one, so we have to update the documentation to reflect this change.

---
 docs/qcow2-cache.txt | 33 ++++++++++++++++-----------------
 1 file changed, 16 insertions(+), 17 deletions(-)

index 170191a242..8a09a5cc5f 100644
@@ -116,31 +116,30 @@ There are three options available, and all of them take bytes:
 "refcount-cache-size":   maximum size of the refcount block cache
 "cache-size":            maximum size of both caches combined
 
-There are two things that need to be taken into account:
+There are a few things that need to be taken into account:
 
  - Both caches must have a size that is a multiple of the cluster size
    (or the cache entry size: see "Using smaller cache sizes" below).
 
- - If you only set one of the options above, QEMU will automatically
-   adjust the others so that the L2 cache is 4 times bigger than the
-   refcount cache.
+ - The default L2 cache size is 8 clusters or 1MB (whichever is more),
+   and the minimum is 2 clusters (or 2 cache entries, see below).
 
-This means that these options are equivalent:
+ - The default (and minimum) refcount cache size is 4 clusters.
 
-   -drive file=hd.qcow2,l2-cache-size=2097152
-   -drive file=hd.qcow2,refcount-cache-size=524288
-   -drive file=hd.qcow2,cache-size=2621440
+ - If only "cache-size" is specified then QEMU will assign as much
+   memory as possible to the L2 cache before increasing the refcount
+   cache size.
 
-The reason for this 1/4 ratio is to ensure that both caches cover the
-same amount of disk space. Note however that this is only valid with
-the default value of refcount_bits (16). If you are using a different
-value you might want to calculate both cache sizes yourself since QEMU
-will always use the same 1/4 ratio.
+Unlike L2 tables, refcount blocks are not used during normal I/O but
+only during allocations and internal snapshots. In most cases they are
+accessed sequentially (even during random guest I/O) so increasing the
+refcount cache size won't have any measurable effect in performance
+(this can change if you are using internal snapshots, so you may want
+to think about increasing the cache size if you use them heavily).
 
-It's also worth mentioning that there's no strict need for both caches
-to cover the same amount of disk space. The refcount cache is used
-much less often than the L2 cache, so it's perfectly reasonable to
-keep it small.
+Before QEMU 2.12 the refcount cache had a default size of 1/4 of the
+L2 cache size. This resulted in unnecessarily large caches, so now the
+refcount cache is as small as possible unless overridden by the user.
 
 
 Using smaller cache entries

2.11.0

The L2 and refcount caches have default sizes that can be overridden using the l2-cache-size and refcount-cache-size (an additional parameter named cache-size sets the combined size of both caches).

Unless forced by one of the aforementioned parameters, QEMU will set the unspecified sizes so that the L2 cache is 4 times larger than the refcount cache.

This is based on the premise that the refcount metadata needs to be only a fourth of the L2 metadata to cover the same amount of disk space. This is incorrect for two reasons:

a) The amount of disk covered by an L2 table depends solely on the cluster size, but in the case of a refcount block it depends on the cluster size *and* the width of each refcount entry. The 4/1 ratio is only valid with 16-bit entries (the default).

b) When we talk about disk space and L2 tables we are talking about guest space (L2 tables map guest clusters to host clusters), whereas refcount blocks are used for host clusters (including L1/L2 tables and the refcount blocks themselves). On a fully populated (and uncompressed) qcow2 file, image size > virtual size so there are more refcount entries than L2 entries.

Problem (a) could be fixed by adjusting the algorithm to take into account the refcount entry width. Problem (b) could be fixed by increasing a bit the refcount cache size to account for the clusters used for qcow2 metadata.

However this patch takes a completely different approach and instead of keeping a ratio between both cache sizes it assigns as much as possible to the L2 cache and the remainder to the refcount cache.

The reason is that L2 tables are used for every single I/O request from the guest and the effect of increasing the cache is significant and clearly measurable. Refcount blocks are however only used for cluster allocation and internal snapshots and in practice are accessed sequentially in most cases, so the effect of increasing the cache is negligible (even when doing random writes from the guest).

So, make the refcount cache as small as possible unless the user explicitly asks for a larger one.

---
 block/qcow2.c              | 31 +++++++++++++++++++------------
 block/qcow2.h              |  4 ----
 tests/qemu-iotests/137.out |  2 +-
 3 files changed, 20 insertions(+), 17 deletions(-)

index ef68772aca..091088e09e 100644
@@ -802,23 +802,30 @@ static void read_cache_sizes(BlockDriverState *bs, QemuOpts *opts,
         } else if (refcount_cache_size_set) {
             *l2_cache_size = combined_cache_size - *refcount_cache_size;
         } else {
-            *refcount_cache_size = combined_cache_size
-                                 / (DEFAULT_L2_REFCOUNT_SIZE_RATIO + 1);
-            *l2_cache_size = combined_cache_size - *refcount_cache_size;
+            uint64_t virtual_disk_size = bs->total_sectors * BDRV_SECTOR_SIZE;
+            uint64_t max_l2_cache = virtual_disk_size / (s->cluster_size / 8);
+            uint64_t min_refcount_cache =
+                (uint64_t) MIN_REFCOUNT_CACHE_SIZE * s->cluster_size;
+
+            /* Assign as much memory as possible to the L2 cache, and
+             * use the remainder for the refcount cache */
+            if (combined_cache_size >= max_l2_cache + min_refcount_cache) {
+                *l2_cache_size = max_l2_cache;
+                *refcount_cache_size = combined_cache_size - *l2_cache_size;
+            } else {
+                *refcount_cache_size =
+                    MIN(combined_cache_size, min_refcount_cache);
+                *l2_cache_size = combined_cache_size - *refcount_cache_size;
+            }
         }
     } else {
-        if (!l2_cache_size_set && !refcount_cache_size_set) {
+        if (!l2_cache_size_set) {
             *l2_cache_size = MAX(DEFAULT_L2_CACHE_BYTE_SIZE,
                                  (uint64_t)DEFAULT_L2_CACHE_CLUSTERS
                                  * s->cluster_size);
-            *refcount_cache_size = *l2_cache_size
-                                 / DEFAULT_L2_REFCOUNT_SIZE_RATIO;
-        } else if (!l2_cache_size_set) {
-            *l2_cache_size = *refcount_cache_size
-                           * DEFAULT_L2_REFCOUNT_SIZE_RATIO;
-        } else if (!refcount_cache_size_set) {
-            *refcount_cache_size = *l2_cache_size
-                                 / DEFAULT_L2_REFCOUNT_SIZE_RATIO;
+        }
+        if (!refcount_cache_size_set) {
+            *refcount_cache_size = MIN_REFCOUNT_CACHE_SIZE * s->cluster_size;
         }
     }
 

index adf5c3950f..01b5250415 100644
@@ -77,10 +77,6 @@
 #define DEFAULT_L2_CACHE_CLUSTERS 8 /* clusters */
 #define DEFAULT_L2_CACHE_BYTE_SIZE 1048576 /* bytes */
 
-/* The refblock cache needs only a fourth of the L2 cache size to cover as many
- * clusters */
-#define DEFAULT_L2_REFCOUNT_SIZE_RATIO 4
-
 #define DEFAULT_CLUSTER_SIZE 65536
 
 

index e28e1eadba..96724a6c33 100644
@@ -22,7 +22,7 @@ refcount-cache-size may not exceed cache-size
 L2 cache size too big
 L2 cache entry size must be a power of two between 512 and the cluster size (65536)
 L2 cache entry size must be a power of two between 512 and the cluster size (65536)
-L2 cache size too big
+Refcount cache size too big
 Conflicting values for qcow2 options 'overlap-check' ('constant') and 'overlap-check.template' ('all')
 Unsupported value 'blubb' for qcow2 option 'overlap-check'. Allowed are any of the following: none, constant, cached, all
 Unsupported value 'blubb' for qcow2 option 'overlap-check'. Allowed are any of the following: none, constant, cached, all

2.11.0

Thanks, applied to my block-next branch:

https://github.com/XanClic/qemu/commits/block-next

Max