distibutions: Extend flexibility of non-uniform random distributions

This change affects options random_distribution and file_service_type.
For pareto, zipf and gauss distribution a contept of `center` is implemented.
It allows to fix in place a value that is most probable to access.
Example:
fio --randseed=1 --ioengine=libaio --rw=randwrite --nrfiles=16 --bs=4k \
    --size=256m --allow_file_create=1 --write_iolog=log.txt \
    --file_service_type=gauss:10:0.1 --filename_format=object.\$filenum --name=x

cat log.txt |grep write |cut -f 1 -d " " |sort |uniq -c | sort -n | \
sed "s/[.]/ /" | while read a b c; do echo $c $b $a; done |sort -n
0 object 13429
1 object 17928
2 object 14724
3 object 7845
4 object 2476
5 object 468
6 object 44
7 object 3
12 object 24
13 object 318
14 object 1795
15 object 6482

Signed-off-by: Adam Kupczyk <[email protected]>

HOWTO

@@ -1361,7 +1361,7 @@ I/O type
 	limit reads or writes to a certain rate.  If that is the case, then the
 	distribution may be skewed. Default: 50.
 
-.. option:: random_distribution=str:float[,str:float][,str:float]
+.. option:: random_distribution=str:float[:float][,str:float][,str:float]
 
 	By default, fio will use a completely uniform random distribution when asked
 	to perform random I/O. Sometimes it is useful to skew the distribution in
@@ -1396,6 +1396,14 @@ I/O type
 	map. For the **normal** distribution, a normal (Gaussian) deviation is
 	supplied as a value between 0 and 100.
 
+	The second, optional float is allowed for **pareto**, **zipf** and **normal** distributions.
+	It allows to set base of distribution in non-default place, giving more control
+	over most probable outcome. This value is in range [0-1] which maps linearly to
+	range of possible random values.
+	Defaults are: random for **pareto** and **zipf**, and 0.5 for **normal**.
+	If you wanted to use **zipf** with a `theta` of 1.2 centered on 1/4 of allowed value range,
+	you would use ``random_distibution=zipf:1.2:0.25``.
+
 	For a **zoned** distribution, fio supports specifying percentages of I/O
 	access that should fall within what range of the file or device. For
 	example, given a criteria of:

cconv.c

@@ -203,6 +203,7 @@ void convert_thread_options_to_cpu(struct thread_options *o,
 	o->zipf_theta.u.f = fio_uint64_to_double(le64_to_cpu(top->zipf_theta.u.i));
 	o->pareto_h.u.f = fio_uint64_to_double(le64_to_cpu(top->pareto_h.u.i));
 	o->gauss_dev.u.f = fio_uint64_to_double(le64_to_cpu(top->gauss_dev.u.i));
+	o->random_center.u.f = fio_uint64_to_double(le64_to_cpu(top->random_center.u.i));
 	o->random_generator = le32_to_cpu(top->random_generator);
 	o->hugepage_size = le32_to_cpu(top->hugepage_size);
 	o->rw_min_bs = le64_to_cpu(top->rw_min_bs);
@@ -423,6 +424,7 @@ void convert_thread_options_to_net(struct thread_options_pack *top,
 	top->zipf_theta.u.i = __cpu_to_le64(fio_double_to_uint64(o->zipf_theta.u.f));
 	top->pareto_h.u.i = __cpu_to_le64(fio_double_to_uint64(o->pareto_h.u.f));
 	top->gauss_dev.u.i = __cpu_to_le64(fio_double_to_uint64(o->gauss_dev.u.f));
+	top->random_center.u.i = __cpu_to_le64(fio_double_to_uint64(o->random_center.u.f));
 	top->random_generator = cpu_to_le32(o->random_generator);
 	top->hugepage_size = cpu_to_le32(o->hugepage_size);
 	top->rw_min_bs = __cpu_to_le64(o->rw_min_bs);

filesetup.c

@@ -1319,11 +1319,11 @@ static void __init_rand_distribution(struct thread_data *td, struct fio_file *f)
 		seed = td->rand_seeds[4];
 
 	if (td->o.random_distribution == FIO_RAND_DIST_ZIPF)
-		zipf_init(&f->zipf, nranges, td->o.zipf_theta.u.f, seed);
+		zipf_init(&f->zipf, nranges, td->o.zipf_theta.u.f, td->o.random_center.u.f, seed);
 	else if (td->o.random_distribution == FIO_RAND_DIST_PARETO)
-		pareto_init(&f->zipf, nranges, td->o.pareto_h.u.f, seed);
+		pareto_init(&f->zipf, nranges, td->o.pareto_h.u.f, td->o.random_center.u.f, seed);
 	else if (td->o.random_distribution == FIO_RAND_DIST_GAUSS)
-		gauss_init(&f->gauss, nranges, td->o.gauss_dev.u.f, seed);
+		gauss_init(&f->gauss, nranges, td->o.gauss_dev.u.f, td->o.random_center.u.f, seed);
 }
 
 static bool init_rand_distribution(struct thread_data *td)

fio.1

@@ -1132,7 +1132,7 @@ first. This may interfere with a given rate setting, if fio is asked to
 limit reads or writes to a certain rate. If that is the case, then the
 distribution may be skewed. Default: 50.
 .TP
-.BI random_distribution \fR=\fPstr:float[,str:float][,str:float]
+.BI random_distribution \fR=\fPstr:float[:float][,str:float][,str:float]
 By default, fio will use a completely uniform random distribution when asked
 to perform random I/O. Sometimes it is useful to skew the distribution in
 specific ways, ensuring that some parts of the data is more hot than others.
@@ -1168,6 +1168,14 @@ option. If a non\-uniform model is used, fio will disable use of the random
 map. For the \fBnormal\fR distribution, a normal (Gaussian) deviation is
 supplied as a value between 0 and 100.
 .P
+The second, optional float is allowed for \fBpareto\fR, \fBzipf\fR and \fBnormal\fR
+distributions. It allows to set base of distribution in non-default place, giving
+more control over most probable outcome. This value is in range [0-1] which maps linearly to
+range of possible random values.
+Defaults are: random for \fBpareto\fR and \fBzipf\fR, and 0.5 for \fBnormal\fR.
+If you wanted to use \fBzipf\fR with a `theta` of 1.2 centered on 1/4 of allowed value range,
+you would use `random_distibution=zipf:1.2:0.25`.
+.P
 For a \fBzoned\fR distribution, fio supports specifying percentages of I/O
 access that should fall within what range of the file or device. For
 example, given a criteria of:

fio.h

@@ -229,6 +229,7 @@ struct thread_data {
 		double pareto_h;
 		double gauss_dev;
 	};
+	double random_center;
 	int error;
 	int sig;
 	int done;

init.c

@@ -971,13 +971,13 @@ static void init_rand_file_service(struct thread_data *td)
 	const unsigned int seed = td->rand_seeds[FIO_RAND_FILE_OFF];
 
 	if (td->o.file_service_type == FIO_FSERVICE_ZIPF) {
-		zipf_init(&td->next_file_zipf, nranges, td->zipf_theta, seed);
+		zipf_init(&td->next_file_zipf, nranges, td->zipf_theta, td->random_center, seed);
 		zipf_disable_hash(&td->next_file_zipf);
 	} else if (td->o.file_service_type == FIO_FSERVICE_PARETO) {
-		pareto_init(&td->next_file_zipf, nranges, td->pareto_h, seed);
+		pareto_init(&td->next_file_zipf, nranges, td->pareto_h, td->random_center, seed);
 		zipf_disable_hash(&td->next_file_zipf);
 	} else if (td->o.file_service_type == FIO_FSERVICE_GAUSS) {
-		gauss_init(&td->next_file_gauss, nranges, td->gauss_dev, seed);
+		gauss_init(&td->next_file_gauss, nranges, td->gauss_dev, td->random_center, seed);
 		gauss_disable_hash(&td->next_file_gauss);
 	}
 }

lib/gauss.c

@@ -40,11 +40,11 @@ unsigned long long gauss_next(struct gauss_state *gs)
 	if (!gs->disable_hash)
 		sum = __hash_u64(sum);
 
-	return sum % gs->nranges;
+	return (sum + gs->rand_off) % gs->nranges;
 }
 
 void gauss_init(struct gauss_state *gs, unsigned long nranges, double dev,
-		unsigned int seed)
+		double center, unsigned int seed)
 {
 	memset(gs, 0, sizeof(*gs));
 	init_rand_seed(&gs->r, seed, 0);
@@ -55,6 +55,10 @@ void gauss_init(struct gauss_state *gs, unsigned long nranges, double dev,
 		if (gs->stddev > nranges / 2)
 			gs->stddev = nranges / 2;
 	}
+	if (center == -1)
+	  gs->rand_off = 0;
+	else
+	  gs->rand_off = nranges * (center - 0.5);
 }
 
 void gauss_disable_hash(struct gauss_state *gs)

lib/gauss.h

@@ -8,11 +8,12 @@ struct gauss_state {
 	struct frand_state r;
 	uint64_t nranges;
 	unsigned int stddev;
+	unsigned int rand_off;
 	bool disable_hash;
 };
 
 void gauss_init(struct gauss_state *gs, unsigned long nranges, double dev,
-		unsigned int seed);
+		double center, unsigned int seed);
 unsigned long long gauss_next(struct gauss_state *gs);
 void gauss_disable_hash(struct gauss_state *gs);
 

lib/zipf.c

@@ -23,19 +23,21 @@ static void zipf_update(struct zipf_state *zs)
 }
 
 static void shared_rand_init(struct zipf_state *zs, uint64_t nranges,
-			     unsigned int seed)
+			     double center, unsigned int seed)
 {
 	memset(zs, 0, sizeof(*zs));
 	zs->nranges = nranges;
 
 	init_rand_seed(&zs->rand, seed, 0);
 	zs->rand_off = __rand(&zs->rand);
+	if (center != -1)
+		zs->rand_off = nranges * center;
 }
 
 void zipf_init(struct zipf_state *zs, uint64_t nranges, double theta,
-	       unsigned int seed)
+	       double center, unsigned int seed)
 {
-	shared_rand_init(zs, nranges, seed);
+	shared_rand_init(zs, nranges, center, seed);
 
 	zs->theta = theta;
 	zs->zeta2 = pow(1.0, zs->theta) + pow(0.5, zs->theta);
@@ -71,9 +73,9 @@ uint64_t zipf_next(struct zipf_state *zs)
 }
 
 void pareto_init(struct zipf_state *zs, uint64_t nranges, double h,
-		 unsigned int seed)
+		 double center, unsigned int seed)
 {
-	shared_rand_init(zs, nranges, seed);
+	shared_rand_init(zs, nranges, center, seed);
 	zs->pareto_pow = log(h) / log(1.0 - h);
 }
 

lib/zipf.h

@@ -16,10 +16,12 @@ struct zipf_state {
 	bool disable_hash;
 };
 
-void zipf_init(struct zipf_state *zs, uint64_t nranges, double theta, unsigned int seed);
+void zipf_init(struct zipf_state *zs, uint64_t nranges, double theta,
+	       double center, unsigned int seed);
 uint64_t zipf_next(struct zipf_state *zs);
 
-void pareto_init(struct zipf_state *zs, uint64_t nranges, double h, unsigned int seed);
+void pareto_init(struct zipf_state *zs, uint64_t nranges, double h,
+		 double center, unsigned int seed);
 uint64_t pareto_next(struct zipf_state *zs);
 void zipf_disable_hash(struct zipf_state *zs);
 

options.c

@@ -44,6 +44,27 @@ static char *get_opt_postfix(const char *str)
 	return strdup(p);
 }
 
+static bool split_parse_distr(const char *str, double *val, double *center)
+{
+	char *cp, *p;
+	bool r;
+
+	p = strdup(str);
+	if (!p)
+		return false;
+
+	cp = strstr(p, ":");
+	r = true;
+	if (cp) {
+		*cp = '\0';
+		cp++;
+		r = str_to_float(cp, center, 0);
+	}
+	r = r && str_to_float(p, val, 0);
+	free(p);
+	return r;
+}
+
 static int bs_cmp(const void *p1, const void *p2)
 {
 	const struct bssplit *bsp1 = p1;
@@ -787,6 +808,7 @@ static int str_fst_cb(void *data, const char *str)
 {
 	struct thread_data *td = cb_data_to_td(data);
 	double val;
+	double center = -1;
 	bool done = false;
 	char *nr;
 
@@ -821,14 +843,20 @@ static int str_fst_cb(void *data, const char *str)
 		return 0;
 
 	nr = get_opt_postfix(str);
-	if (nr && !str_to_float(nr, &val, 0)) {
+	if (nr && !split_parse_distr(nr, &val, &center)) {
 		log_err("fio: file service type random postfix parsing failed\n");
 		free(nr);
 		return 1;
 	}
 
 	free(nr);
 
+	if (center != -1 && (center < 0.00 || center > 1.00)) {
+		log_err("fio: distribution center out of range (0 <= center <= 1.0)\n");
+		return 1;
+	}
+	td->random_center = center;
+
 	switch (td->o.file_service_type) {
 	case FIO_FSERVICE_ZIPF:
 		if (val == 1.00) {
@@ -1030,6 +1058,7 @@ static int str_random_distribution_cb(void *data, const char *str)
 {
 	struct thread_data *td = cb_data_to_td(data);
 	double val;
+	double center = -1;
 	char *nr;
 
 	if (td->o.random_distribution == FIO_RAND_DIST_ZIPF)
@@ -1046,14 +1075,20 @@ static int str_random_distribution_cb(void *data, const char *str)
 		return 0;
 
 	nr = get_opt_postfix(str);
-	if (nr && !str_to_float(nr, &val, 0)) {
+	if (nr && !split_parse_distr(nr, &val, &center)) {
 		log_err("fio: random postfix parsing failed\n");
 		free(nr);
 		return 1;
 	}
 
 	free(nr);
 
+	if (center != -1 && (center < 0.00 || center > 1.00)) {
+		log_err("fio: distribution center out of range (0 <= center <= 1.0)\n");
+		return 1;
+	}
+	td->o.random_center.u.f = center;
+
 	if (td->o.random_distribution == FIO_RAND_DIST_ZIPF) {
 		if (val == 1.00) {
 			log_err("fio: zipf theta must different than 1.0\n");

server.h

@@ -48,7 +48,7 @@ struct fio_net_cmd_reply {
 };
 
 enum {
-	FIO_SERVER_VER			= 86,
+	FIO_SERVER_VER			= 87,
 
 	FIO_SERVER_MAX_FRAGMENT_PDU	= 1024,
 	FIO_SERVER_MAX_CMD_MB		= 2048,

t/genzipf.c

@@ -297,11 +297,11 @@ int main(int argc, char *argv[])
 	nranges /= block_size;
 
 	if (dist_type == TYPE_ZIPF)
-		zipf_init(&zs, nranges, dist_val, 1);
+		zipf_init(&zs, nranges, dist_val, -1, 1);
 	else if (dist_type == TYPE_PARETO)
-		pareto_init(&zs, nranges, dist_val, 1);
+		pareto_init(&zs, nranges, dist_val, -1, 1);
 	else
-		gauss_init(&gs, nranges, dist_val, 1);
+		gauss_init(&gs, nranges, dist_val, -1, 1);
 
 	hash_bits = 0;
 	hash_size = nranges;

thread_options.h

@@ -166,6 +166,7 @@ struct thread_options {
 	fio_fp64_t zipf_theta;
 	fio_fp64_t pareto_h;
 	fio_fp64_t gauss_dev;
+	fio_fp64_t random_center;
 
 	unsigned int random_generator;
 
@@ -467,6 +468,7 @@ struct thread_options_pack {
 	fio_fp64_t zipf_theta;
 	fio_fp64_t pareto_h;
 	fio_fp64_t gauss_dev;
+	fio_fp64_t random_center;
 
 	uint32_t random_generator;