ntdb: remove hash table trees.

TDB2 started with a top-level hash of 1024 entries, divided into 128
groups of 8 buckets.  When a bucket filled, the 8 bucket group
expanded into pointers into 8 new 64-entry hash tables.  When these
filled, they expanded in turn, etc.

It's a nice idea to automatically expand the hash tables, but it
doesn't pay off.  Remove it for NTDB.

1) It only beats TDB performance when the database is huge and the
   TDB hashsize is small.  We are about 20% slower on medium-size
   databases (1000 to 10000 records), worse on really small ones.
2) Since we're 64 bits, our hash tables are already twice as expensive
   as TDB.
3) Since our hash function is good, it means that all groups tend to
   fill at the same time, meaning the hash enlarges by a factor of 128
   all at once, leading to a very large database at that point.
4) Our efficiency would improve if we enlarged the top level, but
   that makes our minimum db size even worse: it's already over 8k,
   and jumps to 1M after about 1000 entries!
5) Making the sub group size larger gives a shallower tree, which
   performs better, but makes the "hash explosion" problem worse.
6) The code is complicated, having to handle delete and reshuffling
   groups of hash buckets, and expansion of buckets.
7) We have to handle the case where all the records somehow end up with
   the same hash value, which requires special code to chain records for
   that case.

On the other hand, it would be nice if we didn't degrade as badly as
TDB does when the hash chains get long.

This patch removes the hash-growing code, but instead of chaining like
TDB does when a bucket fills, we point the bucket to an array of
record pointers.  Since each on-disk NTDB pointer contains some hash
bits from the record (we steal the upper 8 bits of the offset), 99.5%
of the time we don't need to load the record to determine if it
matches.  This makes an array of offsets much more cache-friendly than
a linked list.

Here are the times (in ns) for tdb_store of N records, tdb_store of N
records the second time, and a fetch of all N records.  I've also
included the final database size and the smbtorture local.[n]tdb_speed
results.

Benchmark details:
1) Compiled with -O2.
2) assert() was disabled in TDB2 and NTDB.
3) The "optimize fetch" patch was applied to NTDB.

10 runs, using tmpfs (otherwise massive swapping as db hits ~30M,
despite plenty of RAM).

				Insert	Re-ins	Fetch	Size	dbspeed
				(nsec)	(nsec)	(nsec)	(Kb)	(ops/sec)
TDB (10000 hashsize):	
	100 records:		 3882	 3320	1609	   53	203204
	1000 records:		 3651	 3281	1571	  115	218021
	10000 records:		 3404	 3326	1595	  880	202874
	100000 records:		 4317	 3825	2097	 8262	126811
	1000000 records:	11568	11578	9320	77005	 25046

TDB2 (1024 hashsize, expandable):
	100 records:		 3867	 3329	1699	   17	187100	
	1000 records:		 4040	 3249	1639	  154	186255
	10000 records:		 4143	 3300	1695	 1226	185110
	100000 records:		 4481	 3425	1800	17848	163483
	1000000 records:	 4055	 3534	1878   106386	160774

NTDB (8192 hashsize)
	100 records:		 4259	 3376	1692	   82	190852
	1000 records:		 3640	 3275	1566	  130	195106
	10000 records:		 4337	 3438	1614	  773	188362
	100000 records:		 4750	 5165	1746	 9001	169197
	1000000 records:	 4897	 5180	2341	83838	121901

Analysis:
	1) TDB wins on small databases, beating TDB2 by ~15%, NTDB by ~10%.
	2) TDB starts to lose when hash chains get 10 long (fetch 10% slower
	   than TDB2/NTDB).
	3) TDB does horribly when hash chains get 100 long (fetch 4x slower
	   than NTDB, 5x slower than TDB2, insert about 2-3x slower).
	4) TDB2 databases are 40% larger than TDB1.  NTDB is about 15% larger
	   than TDB1

1 files changed, 31 insertions, 71 deletions

diff --git a/lib/ntdb/test/helprun-layout.c b/lib/ntdb/test/helprun-layout.c
index 7f1f9f9d8e..fa6fa29fce 100644
--- a/lib/ntdb/test/helprun-layout.c
+++ b/lib/ntdb/test/helprun-layout.c
@@ -94,13 +94,6 @@ static ntdb_len_t data_record_len(struct tle_used *used)
 	return len;
 }
 
-static ntdb_len_t hashtable_len(struct tle_hashtable *htable)
-{
-	return sizeof(struct ntdb_used_record)
-		+ (sizeof(ntdb_off_t) << NTDB_SUBLEVEL_HASH_BITS)
-		+ htable->extra;
-}
-
 static ntdb_len_t capability_len(struct tle_capability *cap)
 {
 	return sizeof(struct ntdb_capability) + cap->extra;
@@ -128,25 +121,13 @@ static void set_data_record(void *mem, struct ntdb_context *ntdb,
 	struct ntdb_used_record *u = mem;
 
 	set_header(ntdb, u, NTDB_USED_MAGIC, used->key.dsize, used->data.dsize,
-		   used->key.dsize + used->data.dsize + used->extra,
-		   ntdb_hash(ntdb, used->key.dptr, used->key.dsize));
+		   used->key.dsize + used->data.dsize + used->extra);
 	memcpy(u + 1, used->key.dptr, used->key.dsize);
 	memcpy((char *)(u + 1) + used->key.dsize,
 	       used->data.dptr, used->data.dsize);
 	add_zero_pad(u, used->key.dsize + used->data.dsize, used->extra);
 }
 
-static void set_hashtable(void *mem, struct ntdb_context *ntdb,
-			  struct tle_hashtable *htable)
-{
-	struct ntdb_used_record *u = mem;
-	ntdb_len_t len = sizeof(ntdb_off_t) << NTDB_SUBLEVEL_HASH_BITS;
-
-	set_header(ntdb, u, NTDB_HTABLE_MAGIC, 0, len, len + htable->extra, 0);
-	memset(u + 1, 0, len);
-	add_zero_pad(u, len, htable->extra);
-}
-
 static void set_capability(void *mem, struct ntdb_context *ntdb,
 			   struct tle_capability *cap, struct ntdb_header *hdr,
 			   ntdb_off_t last_cap)
@@ -156,7 +137,7 @@ static void set_capability(void *mem, struct ntdb_context *ntdb,
 
 	c->type = cap->type;
 	c->next = 0;
-	set_header(ntdb, &c->hdr, NTDB_CAP_MAGIC, 0, len, len, 0);
+	set_header(ntdb, &c->hdr, NTDB_CAP_MAGIC, 0, len, len);
 
 	/* Append to capability list. */
 	if (!last_cap) {
@@ -175,7 +156,7 @@ static void set_freetable(void *mem, struct ntdb_context *ntdb,
 	memset(ftable, 0, sizeof(*ftable));
 	set_header(ntdb, &ftable->hdr, NTDB_FTABLE_MAGIC, 0,
 			sizeof(*ftable) - sizeof(ftable->hdr),
-			sizeof(*ftable) - sizeof(ftable->hdr), 0);
+			sizeof(*ftable) - sizeof(ftable->hdr));
 
 	if (last_ftable) {
 		ftable = (struct ntdb_freetable *)((char *)hdr + last_ftable);
@@ -197,12 +178,6 @@ static void add_to_freetable(struct ntdb_context *ntdb,
 			NTDB_LOCK_WAIT, false);
 }
 
-static ntdb_off_t hbucket_off(ntdb_off_t group_start, unsigned ingroup)
-{
-	return group_start
-		+ (ingroup % (1 << NTDB_HASH_GROUP_BITS)) * sizeof(ntdb_off_t);
-}
-
 /* Get bits from a value. */
 static uint32_t bits(uint64_t val, unsigned start, unsigned num)
 {
@@ -210,22 +185,24 @@ static uint32_t bits(uint64_t val, unsigned start, unsigned num)
 	return (val >> start) & ((1U << num) - 1);
 }
 
-/* We take bits from the top: that way we can lock whole sections of the hash
- * by using lock ranges. */
-static uint32_t use_bits(uint64_t h, unsigned num, unsigned *used)
+static ntdb_off_t encode_offset(const struct ntdb_context *ntdb,
+				ntdb_off_t new_off, uint32_t hash)
 {
-	*used += num;
-	return bits(h, 64 - *used, num);
+	ntdb_off_t extra;
+
+	assert((new_off & (1ULL << NTDB_OFF_CHAIN_BIT)) == 0);
+	assert((new_off >> (64 - NTDB_OFF_UPPER_STEAL)) == 0);
+	/* We pack extra hash bits into the upper bits of the offset. */
+	extra = bits(hash, ntdb->hash_bits, NTDB_OFF_UPPER_STEAL);
+	extra <<= (64 - NTDB_OFF_UPPER_STEAL);
+
+	return new_off | extra;
 }
 
-static ntdb_off_t encode_offset(ntdb_off_t new_off, unsigned bucket,
-			       uint64_t h)
+static ntdb_off_t hbucket_off(ntdb_len_t idx)
 {
-	return bucket
-		| new_off
-		| ((uint64_t)bits(h, 64 - NTDB_OFF_UPPER_STEAL_EXTRA,
-				  NTDB_OFF_UPPER_STEAL_EXTRA)
-		   << NTDB_OFF_HASH_EXTRA_BIT);
+	return sizeof(struct ntdb_header) + sizeof(struct ntdb_used_record)
+		+ idx * sizeof(ntdb_off_t);
 }
 
 /* FIXME: Our hash table handling here is primitive: we don't expand! */
@@ -233,28 +210,14 @@ static void add_to_hashtable(struct ntdb_context *ntdb,
 			     ntdb_off_t eoff,
 			     NTDB_DATA key)
 {
-	uint64_t h = ntdb_hash(ntdb, key.dptr, key.dsize);
-	ntdb_off_t b_off, group_start;
-	unsigned i, group, in_group;
-	unsigned used = 0;
+	ntdb_off_t b_off;
+	uint32_t h = ntdb_hash(ntdb, key.dptr, key.dsize);
 
-	group = use_bits(h, NTDB_TOPLEVEL_HASH_BITS-NTDB_HASH_GROUP_BITS, &used);
-	in_group = use_bits(h, NTDB_HASH_GROUP_BITS, &used);
+	b_off = hbucket_off(h & ((1 << ntdb->hash_bits)-1));
+	if (ntdb_read_off(ntdb, b_off) != 0)
+		abort();
 
-	group_start = offsetof(struct ntdb_header, hashtable)
-		+ group * (sizeof(ntdb_off_t) << NTDB_HASH_GROUP_BITS);
-
-	for (i = 0; i < (1 << NTDB_HASH_GROUP_BITS); i++) {
-		unsigned bucket = (in_group + i) % (1 << NTDB_HASH_GROUP_BITS);
-
-		b_off = hbucket_off(group_start, bucket);
-		if (ntdb_read_off(ntdb, b_off) == 0) {
-			ntdb_write_off(ntdb, b_off,
-				      encode_offset(eoff, in_group, h));
-			return;
-		}
-	}
-	abort();
+	ntdb_write_off(ntdb, b_off, encode_offset(ntdb, eoff, h));
 }
 
 static struct tle_freetable *find_ftable(struct ntdb_layout *layout, unsigned num)
@@ -277,11 +240,16 @@ struct ntdb_context *ntdb_layout_get(struct ntdb_layout *layout,
 				   union ntdb_attribute *attr)
 {
 	unsigned int i;
-	ntdb_off_t off, len, last_ftable, last_cap;
+	ntdb_off_t off, hdrlen, len, last_ftable, last_cap;
 	char *mem;
 	struct ntdb_context *ntdb;
 
-	off = sizeof(struct ntdb_header);
+	/* Now populate our header, cribbing from a real NTDB header. */
+	ntdb = ntdb_open("layout", NTDB_INTERNAL, O_RDWR, 0, attr);
+
+	off = sizeof(struct ntdb_header) + sizeof(struct ntdb_used_record)
+		+ (sizeof(ntdb_off_t) << ntdb->hash_bits);
+	hdrlen = off;
 
 	/* First pass of layout: calc lengths */
 	for (i = 0; i < layout->num_elems; i++) {
@@ -297,9 +265,6 @@ struct ntdb_context *ntdb_layout_get(struct ntdb_layout *layout,
 		case DATA:
 			len = data_record_len(&e->used);
 			break;
-		case HASHTABLE:
-			len = hashtable_len(&e->hashtable);
-			break;
 		case CAPABILITY:
 			len = capability_len(&e->capability);
 			break;
@@ -312,9 +277,7 @@ struct ntdb_context *ntdb_layout_get(struct ntdb_layout *layout,
 	mem = malloc(off);
 	/* Fill with some weird pattern. */
 	memset(mem, 0x99, off);
-	/* Now populate our header, cribbing from a real NTDB header. */
-	ntdb = ntdb_open("layout", NTDB_INTERNAL, O_RDWR, 0, attr);
-	memcpy(mem, ntdb->file->map_ptr, sizeof(struct ntdb_header));
+	memcpy(mem, ntdb->file->map_ptr, hdrlen);
 
 	/* Mug the ntdb we have to make it use this. */
 	freefn(ntdb->file->map_ptr);
@@ -337,9 +300,6 @@ struct ntdb_context *ntdb_layout_get(struct ntdb_layout *layout,
 		case DATA:
 			set_data_record(mem + e->base.off, ntdb, &e->used);
 			break;
-		case HASHTABLE:
-			set_hashtable(mem + e->base.off, ntdb, &e->hashtable);
-			break;
 		case CAPABILITY:
 			set_capability(mem + e->base.off, ntdb, &e->capability,
 				       (struct ntdb_header *)mem, last_cap);