1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
|
/* Copyright 2022 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#include "tensorflow/core/data/tfdataz_metrics.h"
#include <algorithm>
#include <cstdint>
#include <deque>
#include <memory>
#include <string>
#include <utility>
#include <vector>
#include "absl/time/time.h"
#include "tensorflow/core/platform/env.h"
namespace tensorflow {
namespace data {
ApproximateLatencyEstimator::ApproximateLatencyEstimator(const Env& env)
: env_(env),
last_updated_time_mins_(0),
latency_value_counter_(0),
latency_count_counter_(0),
next_slot_(0) {
for (int i = 0; i < kSlots; ++i) {
latency_value_[i] = 0;
latency_count_[i] = 0;
}
}
void ApproximateLatencyEstimator::AddLatency(const int64_t latency_usec)
TF_LOCKS_EXCLUDED(mu_) {
UpdateRingBuffer();
mutex_lock l(mu_);
latency_value_counter_ += latency_usec;
latency_count_counter_ += 1;
}
void ApproximateLatencyEstimator::UpdateRingBuffer() TF_LOCKS_EXCLUDED(mu_) {
int64_t now_minutes =
absl::ToInt64Minutes(absl::Microseconds(env_.NowMicros()));
mutex_lock l(mu_);
int64_t elapsed_minutes = now_minutes - last_updated_time_mins_;
int64_t minutes_to_update = std::min(elapsed_minutes, kSlots);
for (int i = 0; i < minutes_to_update; ++i) {
latency_value_[next_slot_] = latency_value_counter_;
latency_count_[next_slot_] = latency_count_counter_;
IncrementNextSlot();
}
last_updated_time_mins_ = now_minutes;
}
void ApproximateLatencyEstimator::IncrementNextSlot()
TF_EXCLUSIVE_LOCKS_REQUIRED(mu_) {
next_slot_ = (next_slot_ + 1) % kSlots;
}
int ApproximateLatencyEstimator::PrevSlot(int steps)
TF_EXCLUSIVE_LOCKS_REQUIRED(mu_) {
return (next_slot_ - steps + kSlots) % kSlots;
}
absl::Duration ApproximateLatencyEstimator::GetAverageLatency(Duration duration)
TF_LOCKS_EXCLUDED(mu_) {
UpdateRingBuffer();
mutex_lock l(mu_);
double interval_latency =
static_cast<double>(latency_value_counter_ -
latency_value_[PrevSlot(static_cast<int>(duration))]);
double interval_count =
static_cast<double>(latency_count_counter_ -
latency_count_[PrevSlot(static_cast<int>(duration))]);
if (interval_count == 0) {
return absl::ZeroDuration();
}
return absl::Duration(absl::Microseconds(interval_latency)) / interval_count;
}
TfDatazMetricsCollector::TfDatazMetricsCollector(const Env& env,
IteratorBase* iterator)
: iterator_(iterator), latency_estimator_(env) {}
void TfDatazMetricsCollector::RecordGetNextLatency(
int64_t get_next_latency_usec) {
if (get_next_latency_usec > 0) {
latency_estimator_.AddLatency(get_next_latency_usec);
}
}
absl::Duration TfDatazMetricsCollector::GetAverageLatencyForLastOneMinute() {
return latency_estimator_.GetAverageLatency(
ApproximateLatencyEstimator::Duration::kMinute);
}
absl::Duration TfDatazMetricsCollector::GetAverageLatencyForLastFiveMinutes() {
return latency_estimator_.GetAverageLatency(
ApproximateLatencyEstimator::Duration::kFiveMinutes);
}
absl::Duration TfDatazMetricsCollector::GetAverageLatencyForLastSixtyMinutes() {
return latency_estimator_.GetAverageLatency(
ApproximateLatencyEstimator::Duration::kSixtyMinutes);
}
int64_t TfDatazMetricsCollector::GetIteratorTotalMemoryUsage() {
return iterator_->TotalBufferedBytes();
}
namespace {
static mutex* get_tfdataz_metrics_registry_lock() {
static mutex tfdataz_metrics_registry_lock(LINKER_INITIALIZED);
return &tfdataz_metrics_registry_lock;
}
using TfDatazMetricsCollectors =
absl::flat_hash_set<std::shared_ptr<TfDatazMetricsCollector>>;
TfDatazMetricsCollectors& tfdataz_metric_collectors() {
static auto& collectors = *new TfDatazMetricsCollectors();
return collectors;
}
} // namespace
void TfDatazMetricsRegistry::Register(
std::shared_ptr<TfDatazMetricsCollector> collector) {
mutex_lock l(*get_tfdataz_metrics_registry_lock());
tfdataz_metric_collectors().insert(collector);
}
void TfDatazMetricsRegistry::Deregister(
std::shared_ptr<TfDatazMetricsCollector> collector) {
mutex_lock l(*get_tfdataz_metrics_registry_lock());
tfdataz_metric_collectors().erase(collector);
}
absl::flat_hash_set<std::shared_ptr<TfDatazMetricsCollector>>
TfDatazMetricsRegistry::GetIteratorMetricCollectors() {
mutex_lock l(*get_tfdataz_metrics_registry_lock());
return tfdataz_metric_collectors();
}
} // namespace data
} // namespace tensorflow
|
