WIP: Adaptive compression [was: auto compression]

cmd/dbufstat/dbufstat.in

@@ -358,7 +358,8 @@ def get_compstring(c):
                     "ZIO_COMPRESS_GZIP_6",  "ZIO_COMPRESS_GZIP_7",
                     "ZIO_COMPRESS_GZIP_8",  "ZIO_COMPRESS_GZIP_9",
                     "ZIO_COMPRESS_ZLE",     "ZIO_COMPRESS_LZ4",
-                    "ZIO_COMPRESS_ZSTD",    "ZIO_COMPRESS_FUNCTION"]
+                    "ZIO_COMPRESS_ZSTD",    "ZIO_COMPRESS_ADAPTIVE",
+                    "ZIO_COMPRESS_FUNCTION"]
 
     # If "-rr" option is used, don't convert to string representation
     if raw > 1:

include/sys/Makefile.am

@@ -15,6 +15,7 @@ COMMON_H = \
 	bptree.h \
 	btree.h \
 	bqueue.h \
+	compress_adaptive.h \
 	dataset_kstats.h \
 	dbuf.h \
 	ddt.h \

include/sys/compress_adaptive.h

@@ -0,0 +1,42 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright (c) 2019, n1kl (bunge)
+ */
+
+#ifndef _SYS_ADAPTIVE_COMPRESS_H
+#define	_SYS_ADAPTIVE_COMPRESS_H
+
+#define	COMPRESS_ADAPTIVE_LEVELS 10
+
+#include <sys/spa.h>
+#include <sys/zio.h>
+
+size_t compress_adaptive(zio_t *zio, abd_t *src, void *dst,
+	size_t s_len, enum zio_compress *c, uint8_t level);
+
+void compress_calc_avg_without_zero(uint64_t act, uint64_t *res, int n);
+
+uint64_t compress_calc_Bps(uint64_t byte, hrtime_t delay);
+
+
+#endif /* _SYS_ADAPTIVE_COMPRESS_H */

include/sys/dsl_dataset.h

@@ -488,6 +488,8 @@ boolean_t dsl_dataset_get_uint64_array_feature(dsl_dataset_t *ds,
 void dsl_dataset_activate_redaction(dsl_dataset_t *ds, uint64_t *redact_snaps,
     uint64_t num_redact_snaps, dmu_tx_t *tx);
 
+int dsl_dataset_activate_compress_adaptive(const char *ddname);
+
 #ifdef ZFS_DEBUG
 #define	dprintf_ds(ds, fmt, ...) do { \
 	if (zfs_flags & ZFS_DEBUG_DPRINTF) { \

include/sys/fs/zfs.h

@@ -1130,6 +1130,9 @@ typedef struct vdev_stat_ex {
 	uint64_t vsx_agg_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
 	    [VDEV_RQ_HISTO_BUCKETS];
 
+
+	uint64_t vsx_diskBps[ZIO_TYPES];
+
 } vdev_stat_ex_t;
 
 /*

include/sys/vdev_impl.h

@@ -132,6 +132,7 @@ struct vdev_cache {
 
 typedef struct vdev_queue_class {
 	uint32_t	vqc_active;
+	uint64_t	vqc_queued_size;
 
 	/*
 	 * Sorted by offset or timestamp, depending on if the queue is

include/sys/zio.h

@@ -41,6 +41,7 @@
 #include <sys/avl.h>
 #include <sys/fs/zfs.h>
 #include <sys/zio_impl.h>
+#include <sys/compress_adaptive.h>
 
 #ifdef	__cplusplus
 extern "C" {
@@ -527,6 +528,12 @@ struct zio {
 
 	/* Taskq dispatching state */
 	taskq_ent_t	io_tqent;
+
+	/* stored in pio */
+	uint8_t		io_compress_level;
+	boolean_t	io_compress_adaptive_exploring;
+
+	uint64_t	io_compress_adaptive_Bps[COMPRESS_ADAPTIVE_LEVELS];
 };
 
 enum blk_verify_flag {

include/sys/zio_compress.h

@@ -54,6 +54,7 @@ enum zio_compress {
 	ZIO_COMPRESS_ZLE,
 	ZIO_COMPRESS_LZ4,
 	ZIO_COMPRESS_ZSTD,
+	ZIO_COMPRESS_ADAPTIVE,
 	ZIO_COMPRESS_FUNCTIONS
 };
 

include/zfeature_common.h

@@ -76,6 +76,7 @@ typedef enum spa_feature {
 	SPA_FEATURE_LIVELIST,
 	SPA_FEATURE_DEVICE_REBUILD,
 	SPA_FEATURE_ZSTD_COMPRESS,
+	SPA_FEATURE_COMPRESS_ADAPTIVE,
 	SPA_FEATURES
 } spa_feature_t;
 

lib/libzpool/Makefile.am

@@ -62,6 +62,7 @@ KERNEL_C = \
 	btree.c \
 	bqueue.c \
 	cityhash.c \
+	compress_adaptive.c \
 	dbuf.c \
 	dbuf_stats.c \
 	ddt.c \

man/man5/zpool-features.5

@@ -16,6 +16,7 @@
 .\" Portions Copyright [yyyy] [name of copyright owner]
 .\" Copyright (c) 2019, Klara Inc.
 .\" Copyright (c) 2019, Allan Jude
+.\" Copyright (c) 2019, RubenKelevra
 .TH ZPOOL-FEATURES 5 "Aug 24, 2020" OpenZFS
 .SH NAME
 zpool\-features \- ZFS pool feature descriptions
@@ -417,7 +418,34 @@ in which new features are enabled. This has no user-visible impact,
 but other features may depend on this feature.
 
 This feature becomes \fBactive\fR as soon as it is enabled and will
-never return to being \fBenabled\fB.
+never return to being \fBenabled\fR.
+.RE
+
+.sp
+.ne 2
+.na
+\fB\fBcompress_adaptive\fR\fR
+.ad
+.RS 4n
+.TS
+l l .
+GUID	org.zfsonlinux:compress_adaptive
+READ\-ONLY COMPATIBLE	yes
+DEPENDENCIES	lz4_compress,extensible_dataset
+.TE
+
+\fBcompress_adaptive\fR chooses the compression algorithm out of lz4 and gzip[1-9] 
+to perform with optimal write throughput.
+
+When the \fBcompress_adaptive\fR feature is set to \fBenabled\fR, the
+administrator can turn on \fBcompress_adaptive\fR on any dataset on the
+pool using the \fBzfs\fR(8) command. Please note that doing so will
+immediately activate the \fBcompress_adaptive\fR feature on the underlying
+pool.
+
+This feature becomes \fBactive\fR as soon as it is used on one dataset and will 
+return to being \fBenabled\fR once all filesystems that have ever had their compression set to
++\fBcompress_adaptive\fR are destroyed.
 .RE
 
 .sp

man/man8/zfsprops.8

@@ -778,7 +778,7 @@ Changing this property affects only newly-written data.
 .It Xo
 .Sy compression Ns = Ns Sy on Ns | Ns Sy off Ns | Ns Sy gzip Ns | Ns
 .Sy gzip- Ns Em N Ns | Ns Sy lz4 Ns | Ns Sy lzjb Ns | Ns Sy zle Ns | Ns Sy zstd Ns | Ns
-.Sy zstd- Ns Em N Ns | Ns Sy zstd-fast Ns | Ns Sy zstd-fast- Ns Em N
+.Sy zstd- Ns Em N Ns | Ns Sy zstd-fast Ns | Ns Sy zstd-fast- Ns Em N | Ns Sy adaptive Ns
 .Xc
 Controls the compression algorithm used for this dataset.
 .Pp
@@ -1989,4 +1989,4 @@ Use the
 .Nm zfs Cm inherit
 command to clear a user property.
 If the property is not defined in any parent dataset, it is removed entirely.
-Property values are limited to 8192 bytes.
+Property values are limited to 8192 bytes.

module/zcommon/zfeature_common.c

@@ -27,6 +27,7 @@
  * Copyright (c) 2017, Intel Corporation.
  * Copyright (c) 2019, Klara Inc.
  * Copyright (c) 2019, Allan Jude
+ * Copyright (c) 2019, n1kl (bunge)
  */
 
 #ifndef _KERNEL
@@ -293,6 +294,19 @@ zpool_feature_init(void)
 	    "LZ4 compression algorithm support.",
 	    ZFEATURE_FLAG_ACTIVATE_ON_ENABLE, ZFEATURE_TYPE_BOOLEAN, NULL);
 
+	{
+	static const spa_feature_t compress_adaptive_deps[] = {
+			SPA_FEATURE_LZ4_COMPRESS,
+			SPA_FEATURE_EXTENSIBLE_DATASET,
+			SPA_FEATURE_NONE
+	};
+	zfeature_register(SPA_FEATURE_COMPRESS_ADAPTIVE,
+	    "org.zfsonlinux:compress_adaptive", "compress_adaptive",
+	    "adaptive compression algorithm selection support.",
+	    ZFEATURE_FLAG_PER_DATASET | ZFEATURE_FLAG_READONLY_COMPAT,
+	    ZFEATURE_TYPE_BOOLEAN, compress_adaptive_deps);
+	}
+
 	zfeature_register(SPA_FEATURE_MULTI_VDEV_CRASH_DUMP,
 	    "com.joyent:multi_vdev_crash_dump", "multi_vdev_crash_dump",
 	    "Crash dumps to multiple vdev pools.",

module/zcommon/zfs_prop.c

@@ -208,6 +208,9 @@ zfs_prop_init(void)
 		    ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_500) },
 		{ "zstd-fast-1000",
 		    ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_1000) },
+
+		{ "adaptive",	ZIO_COMPRESS_ADAPTIVE },
+
 		{ NULL }
 	};
 
@@ -414,7 +417,7 @@ zfs_prop_init(void)
 	zprop_register_index(ZFS_PROP_COMPRESSION, "compression",
 	    ZIO_COMPRESS_DEFAULT, PROP_INHERIT,
 	    ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
-	    "on | off | lzjb | gzip | gzip-[1-9] | zle | lz4 | "
+	    "on | off | adaptive  | lzjb | gzip | gzip-[1-9] | zle | lz4 | "
 	    "zstd | zstd-[1-19] | "
 	    "zstd-fast-[1-10,20,30,40,50,60,70,80,90,100,500,1000]",
 	    "COMPRESS", compress_table);

module/zfs/Makefile.in

@@ -19,6 +19,7 @@ $(MODULE)-objs += arc.o
 $(MODULE)-objs += blkptr.o
 $(MODULE)-objs += bplist.o
 $(MODULE)-objs += bpobj.o
+$(MODULE)-objs += compress_adaptive.o
 $(MODULE)-objs += bptree.o
 $(MODULE)-objs += btree.o
 $(MODULE)-objs += bqueue.o