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
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
|
// Copyright (c) 2012 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "net/quic/congestion_control/cubic.h"
#include <math.h>
#include <algorithm>
#include "base/basictypes.h"
#include "base/logging.h"
#include "base/time/time.h"
#include "net/quic/congestion_control/cube_root.h"
#include "net/quic/quic_flags.h"
#include "net/quic/quic_protocol.h"
using std::max;
namespace net {
namespace {
// Constants based on TCP defaults.
// The following constants are in 2^10 fractions of a second instead of ms to
// allow a 10 shift right to divide.
const int kCubeScale = 40; // 1024*1024^3 (first 1024 is from 0.100^3)
// where 0.100 is 100 ms which is the scaling
// round trip time.
const int kCubeCongestionWindowScale = 410;
const uint64 kCubeFactor = (GG_UINT64_C(1) << kCubeScale) /
kCubeCongestionWindowScale;
const uint32 kDefaultNumConnections = 2;
const float kBeta = 0.7f; // Default Cubic backoff factor.
// Additional backoff factor when loss occurs in the concave part of the Cubic
// curve. This additional backoff factor is expected to give up bandwidth to
// new concurrent flows and speed up convergence.
const float kBetaLastMax = 0.85f;
} // namespace
Cubic::Cubic(const QuicClock* clock, QuicConnectionStats* stats)
: clock_(clock),
num_connections_(kDefaultNumConnections),
epoch_(QuicTime::Zero()),
last_update_time_(QuicTime::Zero()),
stats_(stats) {
Reset();
}
void Cubic::SetNumConnections(int num_connections) {
num_connections_ = num_connections;
}
float Cubic::Alpha() const {
// TCPFriendly alpha is described in Section 3.3 of the CUBIC paper. Note that
// beta here is a cwnd multiplier, and is equal to 1-beta from the paper.
// We derive the equivalent alpha for an N-connection emulation as:
const float beta = Beta();
return 3 * num_connections_ * num_connections_ * (1 - beta) / (1 + beta);
}
float Cubic::Beta() const {
// kNConnectionBeta is the backoff factor after loss for our N-connection
// emulation, which emulates the effective backoff of an ensemble of N
// TCP-Reno connections on a single loss event. The effective multiplier is
// computed as:
return (num_connections_ - 1 + kBeta) / num_connections_;
}
void Cubic::Reset() {
epoch_ = QuicTime::Zero(); // Reset time.
last_update_time_ = QuicTime::Zero(); // Reset time.
last_congestion_window_ = 0;
last_max_congestion_window_ = 0;
acked_packets_count_ = 0;
estimated_tcp_congestion_window_ = 0;
origin_point_congestion_window_ = 0;
time_to_origin_point_ = 0;
last_target_congestion_window_ = 0;
}
void Cubic::UpdateCongestionControlStats(QuicPacketCount new_cubic_mode_cwnd,
QuicPacketCount new_reno_mode_cwnd) {
QuicPacketCount highest_new_cwnd = max(new_cubic_mode_cwnd,
new_reno_mode_cwnd);
if (last_congestion_window_ < highest_new_cwnd) {
// cwnd will increase to highest_new_cwnd.
stats_->cwnd_increase_congestion_avoidance +=
highest_new_cwnd - last_congestion_window_;
if (new_cubic_mode_cwnd > new_reno_mode_cwnd) {
// This cwnd increase is due to cubic mode.
stats_->cwnd_increase_cubic_mode +=
new_cubic_mode_cwnd - last_congestion_window_;
}
}
}
QuicPacketCount Cubic::CongestionWindowAfterPacketLoss(
QuicPacketCount current_congestion_window) {
if (current_congestion_window < last_max_congestion_window_) {
// We never reached the old max, so assume we are competing with another
// flow. Use our extra back off factor to allow the other flow to go up.
last_max_congestion_window_ =
static_cast<int>(kBetaLastMax * current_congestion_window);
} else {
last_max_congestion_window_ = current_congestion_window;
}
epoch_ = QuicTime::Zero(); // Reset time.
return static_cast<int>(current_congestion_window * Beta());
}
QuicPacketCount Cubic::CongestionWindowAfterAck(
QuicPacketCount current_congestion_window,
QuicTime::Delta delay_min) {
acked_packets_count_ += 1; // Packets acked.
QuicTime current_time = clock_->ApproximateNow();
// Cubic is "independent" of RTT, the update is limited by the time elapsed.
if (last_congestion_window_ == current_congestion_window &&
(current_time.Subtract(last_update_time_) <= MaxCubicTimeInterval())) {
return max(last_target_congestion_window_,
estimated_tcp_congestion_window_);
}
last_congestion_window_ = current_congestion_window;
last_update_time_ = current_time;
if (!epoch_.IsInitialized()) {
// First ACK after a loss event.
DVLOG(1) << "Start of epoch";
epoch_ = current_time; // Start of epoch.
acked_packets_count_ = 1; // Reset count.
// Reset estimated_tcp_congestion_window_ to be in sync with cubic.
estimated_tcp_congestion_window_ = current_congestion_window;
if (last_max_congestion_window_ <= current_congestion_window) {
time_to_origin_point_ = 0;
origin_point_congestion_window_ = current_congestion_window;
} else {
if (FLAGS_quic_use_std_cbrt) {
time_to_origin_point_ = static_cast<uint32>(
cbrt(kCubeFactor *
(last_max_congestion_window_ - current_congestion_window)));
} else {
// TODO(rjshade): Remove CubeRoot source when removing
// FLAGS_quic_use_std_cbrt.
time_to_origin_point_ =
CubeRoot::Root(kCubeFactor * (last_max_congestion_window_ -
current_congestion_window));
}
origin_point_congestion_window_ =
last_max_congestion_window_;
}
}
// Change the time unit from microseconds to 2^10 fractions per second. Take
// the round trip time in account. This is done to allow us to use shift as a
// divide operator.
int64 elapsed_time =
(current_time.Add(delay_min).Subtract(epoch_).ToMicroseconds() << 10) /
base::Time::kMicrosecondsPerSecond;
int64 offset = time_to_origin_point_ - elapsed_time;
QuicPacketCount delta_congestion_window = (kCubeCongestionWindowScale
* offset * offset * offset) >> kCubeScale;
QuicPacketCount target_congestion_window =
origin_point_congestion_window_ - delta_congestion_window;
DCHECK_LT(0u, estimated_tcp_congestion_window_);
// With dynamic beta/alpha based on number of active streams, it is possible
// for the required_ack_count to become much lower than acked_packets_count_
// suddenly, leading to more than one iteration through the following loop.
while (true) {
// Update estimated TCP congestion_window.
QuicPacketCount required_ack_count = static_cast<QuicPacketCount>(
estimated_tcp_congestion_window_ / Alpha());
if (acked_packets_count_ < required_ack_count) {
break;
}
acked_packets_count_ -= required_ack_count;
estimated_tcp_congestion_window_++;
}
// Update cubic mode and reno mode stats in QuicConnectionStats.
UpdateCongestionControlStats(target_congestion_window,
estimated_tcp_congestion_window_);
// We have a new cubic congestion window.
last_target_congestion_window_ = target_congestion_window;
// Compute target congestion_window based on cubic target and estimated TCP
// congestion_window, use highest (fastest).
if (target_congestion_window < estimated_tcp_congestion_window_) {
target_congestion_window = estimated_tcp_congestion_window_;
}
DVLOG(1) << "Target congestion_window: " << target_congestion_window;
return target_congestion_window;
}
} // namespace net
|
