The following changes since commit 108fea772db5f1dd91e2fb67737e3e0d36827b76:
client/server: fixup "All clients" reporting (2012-11-14 13:09:45 -0700)
are available in the git repository at:
git://git.kernel.dk/fio.git master
Jens Axboe (6):
server: properly reset stat_number in fio_reset_state()
Get rid of -fno-omit-frame-pointer
genzipf: add help/usage screen (-h)
Disable random map automatically if a non-uniform random distribution is given
Document random distribution
Fio 2.0.11
Vincent Kang Fu (1):
t/genzfip: add -c option for csv output
FIO-VERSION-GEN | 2 +-
HOWTO | 19 ++++++
Makefile | 2 +-
fio.1 | 26 ++++++++
init.c | 7 ++
libfio.c | 1 +
os/windows/install.wxs | 2 +-
server.c | 2 -
t/genzipf.c | 163 +++++++++++++++++++++++++++++-------------------
9 files changed, 155 insertions(+), 69 deletions(-)
---
Diff of recent changes:
diff --git a/FIO-VERSION-GEN b/FIO-VERSION-GEN
index cc7eb83..035ddaf 100755
--- a/FIO-VERSION-GEN
+++ b/FIO-VERSION-GEN
@@ -1,7 +1,7 @@
#!/bin/sh
GVF=FIO-VERSION-FILE
-DEF_VER=fio-2.0.10
+DEF_VER=fio-2.0.11
LF='
'
diff --git a/HOWTO b/HOWTO
index 40fe65f..6391b82 100644
--- a/HOWTO
+++ b/HOWTO
@@ -716,6 +716,25 @@ rwmixwrite=int How large a percentage of the mix should be writes. If both
if fio is asked to limit reads or writes to a certain rate.
If that is the case, then the distribution may be skewed.
+random_distribution=str:float By default, fio will use a completely uniform
+ random distribution when asked to perform random IO. Sometimes
+ it is useful to skew the distribution in specific ways,
+ ensuring that some parts of the data is more hot than others.
+ fio includes the following distribution models:
+
+ random Uniform random distribution
+ zipf Zipf distribution
+ pareto Pareto distribution
+
+ When using a zipf or pareto distribution, an input value
+ is also needed to define the access pattern. For zipf, this
+ is the zipf theta. For pareto, it's the pareto power. Fio
+ includes a test program, genzipf, that can be used visualize
+ what the given input values will yield in terms of hit rates.
+ If you wanted to use zipf with a theta of 1.2, you would use
+ random_distribution=zipf:1.2 as the option. If a non-uniform
+ model is used, fio will disable use of the random map.
+
norandommap Normally fio will cover every block of the file when doing
random IO. If this option is given, fio will just get a
new random offset without looking at past io history. This
diff --git a/Makefile b/Makefile
index 3589770..8473238 100644
--- a/Makefile
+++ b/Makefile
@@ -4,7 +4,7 @@ endif
DEBUGFLAGS = -D_FORTIFY_SOURCE=2 -DFIO_INC_DEBUG
CPPFLAGS= -D_GNU_SOURCE -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 \
$(DEBUGFLAGS)
-OPTFLAGS= -O3 -fno-omit-frame-pointer -g $(EXTFLAGS)
+OPTFLAGS= -O3 -g $(EXTFLAGS)
CFLAGS = -std=gnu99 -Wwrite-strings -Wall $(OPTFLAGS)
LIBS = -lm $(EXTLIBS)
PROGS = fio
diff --git a/fio.1 b/fio.1
index 08d6c0f..2db9793 100644
--- a/fio.1
+++ b/fio.1
@@ -580,6 +580,32 @@ overrides the 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
+By default, fio will use a completely uniform random distribution when asked
+to perform random IO. Sometimes it is useful to skew the distribution in
+specific ways, ensuring that some parts of the data is more hot than others.
+Fio includes the following distribution models:
+.RS
+.TP
+.B random
+Uniform random distribution
+.TP
+.B zipf
+Zipf distribution
+.TP
+.B pareto
+Pareto distribution
+.TP
+.RE
+.P
+When using a zipf or pareto distribution, an input value is also needed to
+define the access pattern. For zipf, this is the zipf theta. For pareto,
+it's the pareto power. Fio includes a test program, genzipf, that can be
+used visualize what the given input values will yield in terms of hit rates.
+If you wanted to use zipf with a theta of 1.2, you would use
+random_distribution=zipf:1.2 as the option. If a non-uniform model is used,
+fio will disable use of the random map.
+.TP
.B norandommap
Normally \fBfio\fR will cover every block of the file when doing random I/O. If
this parameter is given, a new offset will be chosen without looking at past
diff --git a/init.c b/init.c
index a682423..7021741 100644
--- a/init.c
+++ b/init.c
@@ -596,6 +596,13 @@ static int fixup_options(struct thread_data *td)
td->o.compress_percentage = 0;
}
+ /*
+ * Using a non-uniform random distribution excludes usage of
+ * a random map
+ */
+ if (td->o.random_distribution != FIO_RAND_DIST_RANDOM)
+ td->o.norandommap = 1;
+
return ret;
}
diff --git a/libfio.c b/libfio.c
index 43e1a61..0cfd7f6 100644
--- a/libfio.c
+++ b/libfio.c
@@ -127,6 +127,7 @@ void reset_fio_state(void)
{
groupid = 0;
thread_number = 0;
+ stat_number = 0;
nr_process = 0;
nr_thread = 0;
done_secs = 0;
diff --git a/os/windows/install.wxs b/os/windows/install.wxs
index e02347e..dbf0e21 100755
--- a/os/windows/install.wxs
+++ b/os/windows/install.wxs
@@ -10,7 +10,7 @@
<Product Id="2BA394F9-0D9E-4597-BB9D-6B18097D64BB"
Codepage="1252" Language="1033"
Manufacturer="fio" Name="fio"
- UpgradeCode="2338A332-5511-43cf-b9BD-5C60496CCFCC" Version="2.0.10">
+ UpgradeCode="2338A332-5511-43cf-b9BD-5C60496CCFCC" Version="2.0.11">
<Package
Comments="Contact: Your local administrator"
Description="Flexible IO Tester"
diff --git a/server.c b/server.c
index 33b80d6..f8c3635 100644
--- a/server.c
+++ b/server.c
@@ -336,8 +336,6 @@ static int handle_job_cmd(struct fio_net_cmd *cmd)
struct cmd_end_pdu epdu;
int ret;
- stat_number = 0;
-
if (parse_jobs_ini(buf, 1, 0)) {
fio_server_send_quit_cmd();
return -1;
diff --git a/t/genzipf.c b/t/genzipf.c
index 2d1b107..c5f098c 100644
--- a/t/genzipf.c
+++ b/t/genzipf.c
@@ -21,7 +21,6 @@
#include "../lib/zipf.h"
#include "../flist.h"
#include "../hash.h"
-#include "../rbtree.h"
#define DEF_NR 1000000
#define DEF_NR_OUTPUT 23
@@ -49,6 +48,7 @@ static unsigned long block_size = 4096;
static unsigned long output_nranges = DEF_NR_OUTPUT;
static double percentage;
static double dist_val;
+static int output_csv = 0;
#define DEF_ZIPF_VAL 1.2
#define DEF_PARETO_VAL 0.3
@@ -78,13 +78,29 @@ static struct node *hash_insert(struct node *n, unsigned long long val)
return n;
}
+static void usage(void)
+{
+ printf("genzipf: test zipf/pareto values for fio input\n");
+ printf("\t-h\tThis help screen\n");
+ printf("\t-p\tGenerate size of data set that are hit by this percentage\n");
+ printf("\t-t\tDistribution type (zipf or pareto)\n");
+ printf("\t-i\tDistribution algorithm input (zipf theta or pareto power)\n");
+ printf("\t-b\tBlock size of a given range (in bytes)\n");
+ printf("\t-g\tSize of data set (in gigabytes)\n");
+ printf("\t-o\tNumber of output columns\n");
+ printf("\t-c\tOutput ranges in CSV format\n");
+}
+
static int parse_options(int argc, char *argv[])
{
- const char *optstring = "t:g:i:o:b:p:";
+ const char *optstring = "t:g:i:o:b:p:ch";
int c, dist_val_set = 0;
while ((c = getopt(argc, argv, optstring)) != -1) {
switch (c) {
+ case 'h':
+ usage();
+ return 1;
case 'p':
percentage = atof(optarg);
break;
@@ -111,6 +127,9 @@ static int parse_options(int argc, char *argv[])
case 'o':
output_nranges = strtoul(optarg, NULL, 10);
break;
+ case 'c':
+ output_csv = 1;
+ break;
default:
printf("bad option %c\n", c);
return 1;
@@ -162,7 +181,8 @@ int main(int argc, char *argv[])
if (parse_options(argc, argv))
return 1;
- printf("Generating %s distribution with %f input and %lu GB size and %lu block_size.\n", dist_types[dist_type], dist_val, gb_size, block_size);
+ if( !output_csv )
+ printf("Generating %s distribution with %f input and %lu GB size and %lu block_size.\n", dist_types[dist_type], dist_val, gb_size, block_size);
nranges = gb_size * 1024 * 1024 * 1024ULL;
nranges /= block_size;
@@ -208,80 +228,95 @@ int main(int argc, char *argv[])
nnodes = j;
nr_vals = nnodes;
- interval = (nr_vals + output_nranges - 1) / output_nranges;
-
- output_sums = malloc(output_nranges * sizeof(struct output_sum));
- for (i = 0; i < output_nranges; i++) {
- output_sums[i].output = 0.0;
- output_sums[i].nranges = 1;
- }
-
- total_vals = i = j = cur_vals = 0;
-
- for (k = 0; k < nnodes; k++) {
- struct output_sum *os = &output_sums[j];
- struct node *node = &nodes[k];
-
- if (i >= interval) {
- os->output = (double) (cur_vals + 1) / (double) nranges;
- os->output *= 100.0;
- j++;
- cur_vals = node->hits;
- interval += (nr_vals + output_nranges - 1) / output_nranges;
- } else {
- cur_vals += node->hits;
- os->nranges += node->hits;
+ if (output_csv) {
+ printf("rank, count\n");
+ for (k = 0; k < nnodes; k++)
+ printf("%lu, %lu\n", k, nodes[k].hits);
+ } else {
+ interval = (nr_vals + output_nranges - 1) / output_nranges;
+
+ output_sums = malloc(output_nranges * sizeof(struct output_sum));
+ for (i = 0; i < output_nranges; i++) {
+ output_sums[i].output = 0.0;
+ output_sums[i].nranges = 1;
}
- i++;
- total_vals += node->hits;
-
- if (percentage) {
- unsigned long blocks = percentage * nranges / 100;
-
- if (total_vals >= blocks) {
- double cs = i * block_size / (1024 * 1024);
- char p = 'M';
+ total_vals = i = j = cur_vals = 0;
+
+ for (k = 0; k < nnodes; k++) {
+ struct output_sum *os = &output_sums[j];
+ struct node *node = &nodes[k];
+
+ if (i >= interval) {
+ os->output =
+ (double)(cur_vals + 1) / (double)nranges;
+ os->output *= 100.0;
+ j++;
+ cur_vals = node->hits;
+ interval +=
+ (nr_vals + output_nranges -
+ 1) / output_nranges;
+ } else {
+ cur_vals += node->hits;
+ os->nranges += node->hits;
+ }
- if (cs > 1024.0) {
- cs /= 1024.0;
- p = 'G';
+ i++;
+ total_vals += node->hits;
+
+ if (percentage) {
+ unsigned long blocks =
+ percentage * nranges / 100;
+
+ if (total_vals >= blocks) {
+ double cs =
+ i * block_size / (1024 * 1024);
+ char p = 'M';
+
+ if (cs > 1024.0) {
+ cs /= 1024.0;
+ p = 'G';
+ }
+ if (cs > 1024.0) {
+ cs /= 1024.0;
+ p = 'T';
+ }
+
+ printf("%.2f%% of hits satisfied in %.3f%cB of cache\n", percentage, cs, p);
+ percentage = 0.0;
}
- if (cs > 1024.0) {
- cs /= 1024.0;
- p = 'T';
- }
-
- printf("%.2f%% of hits satisfied in %.3f%cB of cache\n", percentage, cs, p);
- percentage = 0.0;
}
}
- }
- perc_i = 100.0 / (double) output_nranges;
- perc = 0.0;
+ perc_i = 100.0 / (double)output_nranges;
+ perc = 0.0;
- printf("\n Rows Hits No Hits Size\n");
- printf("--------------------------------------------------------\n");
- for (i = 0; i < j; i++) {
- struct output_sum *os = &output_sums[i];
- double gb = (double) os->nranges * block_size / 1024.0;
- char p = 'K';
+ printf("\n Rows Hits No Hits Size\n");
+ printf("--------------------------------------------------------\n");
+ for (i = 0; i < j; i++) {
+ struct output_sum *os = &output_sums[i];
+ double gb = (double)os->nranges * block_size / 1024.0;
+ char p = 'K';
- if (gb > 1024.0) {
- p = 'M';
- gb /= 1024.0;
- }
- if (gb > 1024.0) {
- p = 'G';
- gb /= 1024.0;
+ if (gb > 1024.0) {
+ p = 'M';
+ gb /= 1024.0;
+ }
+ if (gb > 1024.0) {
+ p = 'G';
+ gb /= 1024.0;
+ }
+
+ perc += perc_i;
+ printf("%s %6.2f%%\t%6.2f%%\t\t%8u\t%6.2f%c\n",
+ i ? "|->" : "Top", perc, os->output, os->nranges,
+ gb, p);
}
- perc += perc_i;
- printf("%s %6.2f%%\t%6.2f%%\t\t%8u\t%6.2f%c\n", i ? "|->" : "Top", perc, os->output, os->nranges, gb, p);
+ free(output_sums);
}
- free(output_sums);
free(hash);
+ free(nodes);
return 0;
}
--
To unsubscribe from this list: send the line "unsubscribe fio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
