EX-7601 osc: calculate compressed size reduction accurately

Compression reduces space used if it results in allocating
at least one fewer block on disk.  Modify the checks in
compress_chunk to reflect this, rather than using the
simpler "reduce size by at least 4K" calculation.

Also do not attempt to compress chunks if they are less
than 4K in size, since they can't possibly get a space
benefit.

This improved my measured ratio on a version of the Linux
kernel source data set from 1.24 to 1.56, so this is
significant for datasets with many small files.  (This
version of the source had large incompressible files
removed, to focus on smaller files.  The unmodified data set
would not improve as much.)

Note this is still short of our estimates, so either the
estimate or Lustre still needs adjustment.  TBD.

Signed-off-by: Patrick Farrell <pfarrell@whamcloud.com>
Change-Id: I815706914b88de4f532a674d773769aa3a64d218
Reviewed-on: https://review.whamcloud.com/c/ex/lustre-release/+/53181
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Artem Blagodarenko <ablagodarenko@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

diff --git a/lustre/obdclass/lustre_compr.c b/lustre/obdclass/lustre_compr.c
index a62c5c5..3b85182 100644 (file)
--- a/lustre/obdclass/lustre_compr.c
+++ b/lustre/obdclass/lustre_compr.c
@@ -220,8 +220,18 @@ int compress_chunk(const char *obd_name, struct crypto_comp *cc,
 {
        struct ll_compr_hdr *llch;
        unsigned int len = *out_len - sizeof(*llch);
+       int effective_uncompr_size;
+       int effective_compr_size;
        int rc;
 
+       /* the uncompressed data is shorter than the minimum effective
+        * compressed size, so don't bother compressing
+        */
+       if (in_len <= COMP_GAP) {
+               *out_len = in_len;
+               return 0;
+       }
+
        rc = crypto_comp_compress(cc, in, in_len,
                                  out + sizeof(*llch),
                                  &len);
@@ -232,10 +242,15 @@ int compress_chunk(const char *obd_name, struct crypto_comp *cc,
                return 0;
        }
 
-       if (len + sizeof(*llch) + COMP_GAP > in_len) {
+       /* round the sizes up to the nearest block before comparing */
+       effective_compr_size = round_up(len + sizeof(*llch), COMP_GAP);
+       effective_uncompr_size = round_up(in_len, COMP_GAP);
+
+       if (effective_compr_size >= effective_uncompr_size) {
                CDEBUG(D_SEC,
-                      "Compressed %u + overhead %lu > plain %u, leaving uncompressed\n",
-                      len, sizeof(*llch) + COMP_GAP, in_len);
+                      "Compressed %u + overhead %u > plain %u + overhead %u, leaving uncompressed\n",
+                      len, effective_compr_size - len, in_len,
+                      effective_uncompr_size - in_len);
                *out_len = in_len;
                return 0;
        }