#include "fio.h"
#include "steadystate.h"
-#include "helper_thread.h"
bool steadystate_enabled = false;
+unsigned int ss_check_interval = 1000;
+
+void steadystate_free(struct thread_data *td)
+{
+ free(td->ss.iops_data);
+ free(td->ss.bw_data);
+ td->ss.iops_data = NULL;
+ td->ss.bw_data = NULL;
+}
static void steadystate_alloc(struct thread_data *td)
{
- int i;
+ int intervals = td->ss.dur / (ss_check_interval / 1000L);
- td->ss.bw_data = malloc(td->ss.dur * sizeof(unsigned long));
- td->ss.iops_data = malloc(td->ss.dur * sizeof(unsigned long));
- /* initialize so that it is obvious if the cache is not full in the output */
- for (i = 0; i < td->ss.dur; i++)
- td->ss.iops_data[i] = td->ss.bw_data[i] = 0;
+ td->ss.bw_data = calloc(intervals, sizeof(uint64_t));
+ td->ss.iops_data = calloc(intervals, sizeof(uint64_t));
+
+ td->ss.state |= FIO_SS_DATA;
}
void steadystate_setup(void)
{
- int i, prev_groupid;
- struct thread_data *td, *prev_td;
+ struct thread_data *prev_td;
+ int prev_groupid;
if (!steadystate_enabled)
return;
*/
prev_groupid = -1;
prev_td = NULL;
- for_each_td(td, i) {
- if (td->ts.ss == NULL)
+ for_each_td(td) {
+ if (!td->ss.dur)
continue;
if (!td->o.group_reporting) {
}
if (prev_groupid != td->groupid) {
- if (prev_td != NULL) {
- prev_td->ss.state |= __FIO_SS_LAST;
+ if (prev_td)
steadystate_alloc(prev_td);
- }
prev_groupid = td->groupid;
}
prev_td = td;
- }
+ } end_for_each();
- if (prev_td != NULL && prev_td->o.group_reporting) {
- prev_td->ss.state |= __FIO_SS_LAST;
+ if (prev_td && prev_td->o.group_reporting)
steadystate_alloc(prev_td);
- }
}
-static bool steadystate_slope(unsigned long iops, unsigned long bw,
+static bool steadystate_slope(uint64_t iops, uint64_t bw,
struct thread_data *td)
{
int i, j;
double result;
struct steadystate_data *ss = &td->ss;
- unsigned long new_val;
+ uint64_t new_val;
+ int intervals = ss->dur / (ss_check_interval / 1000L);
ss->bw_data[ss->tail] = bw;
ss->iops_data[ss->tail] = iops;
- if (ss->state & __FIO_SS_IOPS)
+ if (ss->state & FIO_SS_IOPS)
new_val = iops;
else
new_val = bw;
- if (ss->tail < ss->head || (ss->tail - ss->head == ss->dur - 1)) {
- if (ss->sum_y == 0) { /* first time through */
- for(i = 0; i < ss->dur; i++) {
- if (ss->state & __FIO_SS_IOPS)
+ if (ss->state & FIO_SS_BUFFER_FULL || ss->tail - ss->head == intervals - 1) {
+ if (!(ss->state & FIO_SS_BUFFER_FULL)) {
+ /* first time through */
+ for (i = 0, ss->sum_y = 0; i < intervals; i++) {
+ if (ss->state & FIO_SS_IOPS)
ss->sum_y += ss->iops_data[i];
else
ss->sum_y += ss->bw_data[i];
- j = ss->head + i;
- if (j >= ss->dur)
- j -= ss->dur;
- if (ss->state & __FIO_SS_IOPS)
+ j = (ss->head + i) % intervals;
+ if (ss->state & FIO_SS_IOPS)
ss->sum_xy += i * ss->iops_data[j];
else
ss->sum_xy += i * ss->bw_data[j];
}
+ ss->state |= FIO_SS_BUFFER_FULL;
} else { /* easy to update the sums */
ss->sum_y -= ss->oldest_y;
ss->sum_y += new_val;
- ss->sum_xy = ss->sum_xy - ss->sum_y + ss->dur * new_val;
+ ss->sum_xy = ss->sum_xy - ss->sum_y + intervals * new_val;
}
- if (ss->state & __FIO_SS_IOPS)
+ if (ss->state & FIO_SS_IOPS)
ss->oldest_y = ss->iops_data[ss->head];
else
ss->oldest_y = ss->bw_data[ss->head];
* equally spaced when they are often off by a few milliseconds.
* This assumption greatly simplifies the calculations.
*/
- ss->slope = (ss->sum_xy - (double) ss->sum_x * ss->sum_y / ss->dur) /
- (ss->sum_x_sq - (double) ss->sum_x * ss->sum_x / ss->dur);
- if (ss->pct)
- ss->criterion = 100.0 * ss->slope / (ss->sum_y / ss->dur);
+ ss->slope = (ss->sum_xy - (double) ss->sum_x * ss->sum_y / intervals) /
+ (ss->sum_x_sq - (double) ss->sum_x * ss->sum_x / intervals);
+ if (ss->state & FIO_SS_PCT)
+ ss->criterion = 100.0 * ss->slope / (ss->sum_y / intervals);
else
ss->criterion = ss->slope;
dprint(FD_STEADYSTATE, "sum_y: %llu, sum_xy: %llu, slope: %f, "
"criterion: %f, limit: %f\n",
- ss->sum_y, ss->sum_xy, ss->slope,
- ss->criterion, ss->limit);
+ (unsigned long long) ss->sum_y,
+ (unsigned long long) ss->sum_xy,
+ ss->slope, ss->criterion, ss->limit);
result = ss->criterion * (ss->criterion < 0.0 ? -1.0 : 1.0);
if (result < ss->limit)
return true;
}
- ss->tail = (ss->tail + 1) % ss->dur;
+ ss->tail = (ss->tail + 1) % intervals;
if (ss->tail <= ss->head)
- ss->head = (ss->head + 1) % ss->dur;
+ ss->head = (ss->head + 1) % intervals;
return false;
}
-static bool steadystate_deviation(unsigned long iops, unsigned long bw,
+static bool steadystate_deviation(uint64_t iops, uint64_t bw,
struct thread_data *td)
{
int i;
double mean;
struct steadystate_data *ss = &td->ss;
+ int intervals = ss->dur / (ss_check_interval / 1000L);
ss->bw_data[ss->tail] = bw;
ss->iops_data[ss->tail] = iops;
- if (ss->tail < ss->head || (ss->tail - ss->head == ss->dur - 1)) {
- if (ss->sum_y == 0) { /* first time through */
- for(i = 0; i < ss->dur; i++)
- if (ss->state & __FIO_SS_IOPS)
+ if (ss->state & FIO_SS_BUFFER_FULL || ss->tail - ss->head == intervals - 1) {
+ if (!(ss->state & FIO_SS_BUFFER_FULL)) {
+ /* first time through */
+ for (i = 0, ss->sum_y = 0; i < intervals; i++) {
+ if (ss->state & FIO_SS_IOPS)
ss->sum_y += ss->iops_data[i];
else
ss->sum_y += ss->bw_data[i];
+ }
+ ss->state |= FIO_SS_BUFFER_FULL;
} else { /* easy to update the sum */
ss->sum_y -= ss->oldest_y;
- if (ss->state & __FIO_SS_IOPS)
+ if (ss->state & FIO_SS_IOPS)
ss->sum_y += ss->iops_data[ss->tail];
else
ss->sum_y += ss->bw_data[ss->tail];
}
- if (ss->state & __FIO_SS_IOPS)
+ if (ss->state & FIO_SS_IOPS)
ss->oldest_y = ss->iops_data[ss->head];
else
ss->oldest_y = ss->bw_data[ss->head];
- mean = (double) ss->sum_y / ss->dur;
+ mean = (double) ss->sum_y / intervals;
ss->deviation = 0.0;
- for (i = 0; i < ss->dur; i++) {
- if (ss->state & __FIO_SS_IOPS)
+ for (i = 0; i < intervals; i++) {
+ if (ss->state & FIO_SS_IOPS)
diff = ss->iops_data[i] - mean;
else
diff = ss->bw_data[i] - mean;
ss->deviation = max(ss->deviation, diff * (diff < 0.0 ? -1.0 : 1.0));
}
- if (ss->pct)
+ if (ss->state & FIO_SS_PCT)
ss->criterion = 100.0 * ss->deviation / mean;
else
ss->criterion = ss->deviation;
- dprint(FD_STEADYSTATE, "sum_y: %llu, mean: %f, max diff: %f, "
+ dprint(FD_STEADYSTATE, "intervals: %d, sum_y: %llu, mean: %f, max diff: %f, "
"objective: %f, limit: %f\n",
- ss->sum_y, mean, ss->deviation,
- ss->criterion, ss->limit);
+ intervals,
+ (unsigned long long) ss->sum_y, mean,
+ ss->deviation, ss->criterion, ss->limit);
if (ss->criterion < ss->limit)
return true;
}
- ss->tail = (ss->tail + 1) % ss->dur;
- if (ss->tail <= ss->head)
- ss->head = (ss->head + 1) % ss->dur;
+ ss->tail = (ss->tail + 1) % intervals;
+ if (ss->tail == ss->head)
+ ss->head = (ss->head + 1) % intervals;
return false;
}
-void steadystate_check(void)
+int steadystate_check(void)
{
- int i, j, ddir, prev_groupid, group_ramp_time_over = 0;
+ int ddir, prev_groupid, group_ramp_time_over = 0;
unsigned long rate_time;
- struct thread_data *td, *td2;
- struct timeval now;
- unsigned long group_bw = 0, group_iops = 0;
- unsigned long long td_iops;
- unsigned long long td_bytes;
+ struct timespec now;
+ uint64_t group_bw = 0, group_iops = 0;
+ uint64_t td_iops, td_bytes;
bool ret;
prev_groupid = -1;
- for_each_td(td, i) {
+ for_each_td(td) {
+ const bool needs_lock = td_async_processing(td);
struct steadystate_data *ss = &td->ss;
if (!ss->dur || td->runstate <= TD_SETTING_UP ||
- td->runstate >= TD_EXITED || (ss->state & __FIO_SS_ATTAINED))
+ td->runstate >= TD_EXITED || !ss->state ||
+ ss->state & FIO_SS_ATTAINED)
continue;
td_iops = 0;
prev_groupid = td->groupid;
fio_gettime(&now, NULL);
- if (ss->ramp_time && !(ss->state & __FIO_SS_RAMP_OVER)) {
+ if (ss->ramp_time && !(ss->state & FIO_SS_RAMP_OVER)) {
/*
- * Begin recording data one second after ss->ramp_time
+ * Begin recording data one check interval after ss->ramp_time
* has elapsed
*/
- if (utime_since(&td->epoch, &now) >= (ss->ramp_time + 1000000L))
- ss->state |= __FIO_SS_RAMP_OVER;
+ if (utime_since(&td->epoch, &now) >= (ss->ramp_time + ss_check_interval * 1000L))
+ ss->state |= FIO_SS_RAMP_OVER;
}
- td_io_u_lock(td);
- for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
+ if (needs_lock)
+ __td_io_u_lock(td);
+
+ for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++) {
td_iops += td->io_blocks[ddir];
td_bytes += td->io_bytes[ddir];
}
- td_io_u_unlock(td);
+
+ if (needs_lock)
+ __td_io_u_unlock(td);
rate_time = mtime_since(&ss->prev_time, &now);
memcpy(&ss->prev_time, &now, sizeof(now));
- /*
- * Begin monitoring when job starts but don't actually use
- * data in checking stopping criterion until ss->ramp_time is
- * over. This ensures that we will have a sane value in
- * prev_iops/bw the first time through after ss->ramp_time
- * is done.
- */
- if (ss->state & __FIO_SS_RAMP_OVER) {
- group_bw += 1000 * (td_bytes - ss->prev_bytes) / rate_time;
- group_iops += 1000 * (td_iops - ss->prev_iops) / rate_time;
+ if (ss->state & FIO_SS_RAMP_OVER) {
+ group_bw += rate_time * (td_bytes - ss->prev_bytes) /
+ (ss_check_interval * ss_check_interval / 1000L);
+ group_iops += rate_time * (td_iops - ss->prev_iops) /
+ (ss_check_interval * ss_check_interval / 1000L);
++group_ramp_time_over;
}
ss->prev_iops = td_iops;
ss->prev_bytes = td_bytes;
- if (td->o.group_reporting && !(ss->state & __FIO_SS_LAST))
+ if (td->o.group_reporting && !(ss->state & FIO_SS_DATA))
continue;
/*
dprint(FD_STEADYSTATE, "steadystate_check() thread: %d, "
"groupid: %u, rate_msec: %ld, "
- "iops: %lu, bw: %lu, head: %d, tail: %d\n",
- i, td->groupid, rate_time, group_iops,
- group_bw, ss->head, ss->tail);
+ "iops: %llu, bw: %llu, head: %d, tail: %d\n",
+ __td_index, td->groupid, rate_time,
+ (unsigned long long) group_iops,
+ (unsigned long long) group_bw,
+ ss->head, ss->tail);
- if (td->o.ss & __FIO_SS_SLOPE)
+ if (ss->state & FIO_SS_SLOPE)
ret = steadystate_slope(group_iops, group_bw, td);
else
ret = steadystate_deviation(group_iops, group_bw, td);
if (ret) {
if (td->o.group_reporting) {
- for_each_td(td2, j) {
+ for_each_td(td2) {
if (td2->groupid == td->groupid) {
- td2->ss.state |= __FIO_SS_ATTAINED;
+ td2->ss.state |= FIO_SS_ATTAINED;
fio_mark_td_terminate(td2);
}
- }
+ } end_for_each();
} else {
- ss->state |= __FIO_SS_ATTAINED;
+ ss->state |= FIO_SS_ATTAINED;
fio_mark_td_terminate(td);
}
}
- }
+ } end_for_each();
+ return 0;
}
int td_steadystate_init(struct thread_data *td)
{
struct steadystate_data *ss = &td->ss;
struct thread_options *o = &td->o;
- struct thread_data *td2;
- int j;
+ int intervals;
memset(ss, 0, sizeof(*ss));
ss->dur = o->ss_dur;
ss->limit = o->ss_limit.u.f;
ss->ramp_time = o->ss_ramp_time;
- ss->pct = o->ss_pct;
+ ss_check_interval = o->ss_check_interval / 1000L;
- ss->state = o->ss;
+ ss->state = o->ss_state;
if (!td->ss.ramp_time)
- ss->state |= __FIO_SS_RAMP_OVER;
+ ss->state |= FIO_SS_RAMP_OVER;
- ss->sum_x = o->ss_dur * (o->ss_dur - 1) / 2;
- ss->sum_x_sq = (o->ss_dur - 1) * (o->ss_dur) * (2*o->ss_dur - 1) / 6;
-
- td->ts.ss = ss;
+ intervals = ss->dur / (ss_check_interval / 1000L);
+ ss->sum_x = intervals * (intervals - 1) / 2;
+ ss->sum_x_sq = (intervals - 1) * (intervals) * (2*intervals - 1) / 6;
}
/* make sure that ss options are consistent within reporting group */
- for_each_td(td2, j) {
+ for_each_td(td2) {
if (td2->groupid == td->groupid) {
struct steadystate_data *ss2 = &td2->ss;
if (ss2->dur != ss->dur ||
ss2->limit != ss->limit ||
ss2->ramp_time != ss->ramp_time ||
- ss2->pct != ss->pct ||
ss2->state != ss->state ||
ss2->sum_x != ss->sum_x ||
ss2->sum_x_sq != ss->sum_x_sq) {
return 1;
}
}
- }
+ } end_for_each();
return 0;
}
+
+uint64_t steadystate_bw_mean(struct thread_stat *ts)
+{
+ int i;
+ uint64_t sum;
+ int intervals = ts->ss_dur / (ss_check_interval / 1000L);
+
+ if (!ts->ss_dur)
+ return 0;
+
+ for (i = 0, sum = 0; i < intervals; i++)
+ sum += ts->ss_bw_data[i];
+
+ return sum / intervals;
+}
+
+uint64_t steadystate_iops_mean(struct thread_stat *ts)
+{
+ int i;
+ uint64_t sum;
+ int intervals = ts->ss_dur / (ss_check_interval / 1000L);
+
+ if (!ts->ss_dur)
+ return 0;
+
+ for (i = 0, sum = 0; i < intervals; i++)
+ sum += ts->ss_iops_data[i];
+
+ return sum / intervals;
+}