blob: c00b76c1c3daf0c9ee8d86b25d23ddcf6ce82ab4 [file] [log] [blame]
// Copyright (c) 2015 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/tcp_cubic_bytes_sender.h"
#include <algorithm>
#include "base/logging.h"
#include "base/memory/scoped_ptr.h"
#include "net/quic/congestion_control/rtt_stats.h"
#include "net/quic/crypto/crypto_protocol.h"
#include "net/quic/proto/cached_network_parameters.pb.h"
#include "net/quic/quic_protocol.h"
#include "net/quic/quic_utils.h"
#include "net/quic/test_tools/mock_clock.h"
#include "net/quic/test_tools/quic_config_peer.h"
#include "testing/gtest/include/gtest/gtest.h"
namespace net {
namespace test {
// TODO(ianswett): A number of theses tests were written with the assumption of
// an initial CWND of 10. They have carefully calculated values which should be
// updated to be based on kInitialCongestionWindowInsecure.
const uint32 kInitialCongestionWindowPackets = 10;
const uint32 kDefaultWindowTCP =
kInitialCongestionWindowPackets * kDefaultTCPMSS;
const float kRenoBeta = 0.7f; // Reno backoff factor.
class TcpCubicBytesSenderPeer : public TcpCubicBytesSender {
public:
TcpCubicBytesSenderPeer(const QuicClock* clock, bool reno)
: TcpCubicBytesSender(clock,
&rtt_stats_,
reno,
kInitialCongestionWindowPackets,
kMaxTcpCongestionWindow,
&stats_) {}
const HybridSlowStart& hybrid_slow_start() const {
return hybrid_slow_start_;
}
float GetRenoBeta() const { return RenoBeta(); }
RttStats rtt_stats_;
QuicConnectionStats stats_;
};
class TcpCubicBytesSenderTest : public ::testing::Test {
protected:
TcpCubicBytesSenderTest()
: one_ms_(QuicTime::Delta::FromMilliseconds(1)),
sender_(new TcpCubicBytesSenderPeer(&clock_, true)),
sequence_number_(1),
acked_sequence_number_(0),
bytes_in_flight_(0) {
standard_packet_.bytes_sent = kDefaultTCPMSS;
}
int SendAvailableSendWindow() {
// Send as long as TimeUntilSend returns Zero.
int packets_sent = 0;
bool can_send = sender_->TimeUntilSend(clock_.Now(), bytes_in_flight_,
HAS_RETRANSMITTABLE_DATA).IsZero();
while (can_send) {
sender_->OnPacketSent(clock_.Now(), bytes_in_flight_, sequence_number_++,
kDefaultTCPMSS, HAS_RETRANSMITTABLE_DATA);
++packets_sent;
bytes_in_flight_ += kDefaultTCPMSS;
can_send = sender_->TimeUntilSend(clock_.Now(), bytes_in_flight_,
HAS_RETRANSMITTABLE_DATA).IsZero();
}
return packets_sent;
}
// Normal is that TCP acks every other segment.
void AckNPackets(int n) {
sender_->rtt_stats_.UpdateRtt(QuicTime::Delta::FromMilliseconds(60),
QuicTime::Delta::Zero(), clock_.Now());
SendAlgorithmInterface::CongestionVector acked_packets;
SendAlgorithmInterface::CongestionVector lost_packets;
for (int i = 0; i < n; ++i) {
++acked_sequence_number_;
acked_packets.push_back(
std::make_pair(acked_sequence_number_, standard_packet_));
}
sender_->OnCongestionEvent(true, bytes_in_flight_, acked_packets,
lost_packets);
bytes_in_flight_ -= n * kDefaultTCPMSS;
clock_.AdvanceTime(one_ms_);
}
void LoseNPackets(int n) {
SendAlgorithmInterface::CongestionVector acked_packets;
SendAlgorithmInterface::CongestionVector lost_packets;
for (int i = 0; i < n; ++i) {
++acked_sequence_number_;
lost_packets.push_back(
std::make_pair(acked_sequence_number_, standard_packet_));
}
sender_->OnCongestionEvent(false, bytes_in_flight_, acked_packets,
lost_packets);
bytes_in_flight_ -= n * kDefaultTCPMSS;
}
// Does not increment acked_sequence_number_.
void LosePacket(QuicPacketSequenceNumber sequence_number) {
SendAlgorithmInterface::CongestionVector acked_packets;
SendAlgorithmInterface::CongestionVector lost_packets;
lost_packets.push_back(std::make_pair(sequence_number, standard_packet_));
sender_->OnCongestionEvent(false, bytes_in_flight_, acked_packets,
lost_packets);
bytes_in_flight_ -= kDefaultTCPMSS;
}
const QuicTime::Delta one_ms_;
MockClock clock_;
scoped_ptr<TcpCubicBytesSenderPeer> sender_;
QuicPacketSequenceNumber sequence_number_;
QuicPacketSequenceNumber acked_sequence_number_;
QuicByteCount bytes_in_flight_;
TransmissionInfo standard_packet_;
};
TEST_F(TcpCubicBytesSenderTest, SimpleSender) {
// At startup make sure we are at the default.
EXPECT_EQ(kDefaultWindowTCP, sender_->GetCongestionWindow());
// At startup make sure we can send.
EXPECT_TRUE(sender_->TimeUntilSend(clock_.Now(), 0,
HAS_RETRANSMITTABLE_DATA).IsZero());
// Make sure we can send.
EXPECT_TRUE(sender_->TimeUntilSend(clock_.Now(), 0,
HAS_RETRANSMITTABLE_DATA).IsZero());
// And that window is un-affected.
EXPECT_EQ(kDefaultWindowTCP, sender_->GetCongestionWindow());
// Fill the send window with data, then verify that we can't send.
SendAvailableSendWindow();
EXPECT_FALSE(sender_->TimeUntilSend(clock_.Now(),
sender_->GetCongestionWindow(),
HAS_RETRANSMITTABLE_DATA).IsZero());
}
TEST_F(TcpCubicBytesSenderTest, ApplicationLimitedSlowStart) {
// Send exactly 10 packets and ensure the CWND ends at 14 packets.
const int kNumberOfAcks = 5;
// At startup make sure we can send.
EXPECT_TRUE(sender_->TimeUntilSend(clock_.Now(), 0,
HAS_RETRANSMITTABLE_DATA).IsZero());
// Make sure we can send.
EXPECT_TRUE(sender_->TimeUntilSend(clock_.Now(), 0,
HAS_RETRANSMITTABLE_DATA).IsZero());
SendAvailableSendWindow();
for (int i = 0; i < kNumberOfAcks; ++i) {
AckNPackets(2);
}
QuicByteCount bytes_to_send = sender_->GetCongestionWindow();
// It's expected 2 acks will arrive when the bytes_in_flight are greater than
// half the CWND.
EXPECT_EQ(kDefaultWindowTCP + kDefaultTCPMSS * 2 * 2, bytes_to_send);
}
TEST_F(TcpCubicBytesSenderTest, ExponentialSlowStart) {
const int kNumberOfAcks = 20;
// At startup make sure we can send.
EXPECT_TRUE(sender_->TimeUntilSend(clock_.Now(), 0,
HAS_RETRANSMITTABLE_DATA).IsZero());
EXPECT_FALSE(sender_->HasReliableBandwidthEstimate());
EXPECT_EQ(QuicBandwidth::Zero(), sender_->BandwidthEstimate());
// Make sure we can send.
EXPECT_TRUE(sender_->TimeUntilSend(clock_.Now(), 0,
HAS_RETRANSMITTABLE_DATA).IsZero());
for (int i = 0; i < kNumberOfAcks; ++i) {
// Send our full send window.
SendAvailableSendWindow();
AckNPackets(2);
}
const QuicByteCount cwnd = sender_->GetCongestionWindow();
EXPECT_EQ(kDefaultWindowTCP + kDefaultTCPMSS * 2 * kNumberOfAcks, cwnd);
EXPECT_FALSE(sender_->HasReliableBandwidthEstimate());
EXPECT_EQ(QuicBandwidth::FromBytesAndTimeDelta(
cwnd, sender_->rtt_stats_.smoothed_rtt()),
sender_->BandwidthEstimate());
}
TEST_F(TcpCubicBytesSenderTest, SlowStartPacketLoss) {
sender_->SetNumEmulatedConnections(1);
const int kNumberOfAcks = 10;
for (int i = 0; i < kNumberOfAcks; ++i) {
// Send our full send window.
SendAvailableSendWindow();
AckNPackets(2);
}
SendAvailableSendWindow();
QuicByteCount expected_send_window =
kDefaultWindowTCP + (kDefaultTCPMSS * 2 * kNumberOfAcks);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Lose a packet to exit slow start.
LoseNPackets(1);
size_t packets_in_recovery_window = expected_send_window / kDefaultTCPMSS;
// We should now have fallen out of slow start with a reduced window.
expected_send_window *= kRenoBeta;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Recovery phase. We need to ack every packet in the recovery window before
// we exit recovery.
size_t number_of_packets_in_window = expected_send_window / kDefaultTCPMSS;
DVLOG(1) << "number_packets: " << number_of_packets_in_window;
AckNPackets(packets_in_recovery_window);
SendAvailableSendWindow();
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// We need to ack an entire window before we increase CWND by 1.
AckNPackets(number_of_packets_in_window - 2);
SendAvailableSendWindow();
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Next ack should increase cwnd by 1.
AckNPackets(1);
expected_send_window += kDefaultTCPMSS;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Now RTO and ensure slow start gets reset.
EXPECT_TRUE(sender_->hybrid_slow_start().started());
sender_->OnRetransmissionTimeout(true);
EXPECT_FALSE(sender_->hybrid_slow_start().started());
}
TEST_F(TcpCubicBytesSenderTest, NoPRRWhenLessThanOnePacketInFlight) {
SendAvailableSendWindow();
LoseNPackets(kInitialCongestionWindowPackets - 1);
AckNPackets(1);
// PRR will allow 2 packets for every ack during recovery.
EXPECT_EQ(2, SendAvailableSendWindow());
// Simulate abandoning all packets by supplying a bytes_in_flight of 0.
// PRR should now allow a packet to be sent, even though prr's state variables
// believe it has sent enough packets.
EXPECT_EQ(QuicTime::Delta::Zero(),
sender_->TimeUntilSend(clock_.Now(), 0, HAS_RETRANSMITTABLE_DATA));
}
TEST_F(TcpCubicBytesSenderTest, SlowStartPacketLossPRR) {
sender_->SetNumEmulatedConnections(1);
// Test based on the first example in RFC6937.
// Ack 10 packets in 5 acks to raise the CWND to 20, as in the example.
const int kNumberOfAcks = 5;
for (int i = 0; i < kNumberOfAcks; ++i) {
// Send our full send window.
SendAvailableSendWindow();
AckNPackets(2);
}
SendAvailableSendWindow();
QuicByteCount expected_send_window =
kDefaultWindowTCP + (kDefaultTCPMSS * 2 * kNumberOfAcks);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
LoseNPackets(1);
// We should now have fallen out of slow start with a reduced window.
size_t send_window_before_loss = expected_send_window;
expected_send_window *= kRenoBeta;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Testing TCP proportional rate reduction.
// We should send packets paced over the received acks for the remaining
// outstanding packets. The number of packets before we exit recovery is the
// original CWND minus the packet that has been lost and the one which
// triggered the loss.
size_t remaining_packets_in_recovery =
send_window_before_loss / kDefaultTCPMSS - 2;
for (size_t i = 0; i < remaining_packets_in_recovery; ++i) {
AckNPackets(1);
SendAvailableSendWindow();
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
}
// We need to ack another window before we increase CWND by 1.
size_t number_of_packets_in_window = expected_send_window / kDefaultTCPMSS;
for (size_t i = 0; i < number_of_packets_in_window; ++i) {
AckNPackets(1);
EXPECT_EQ(1, SendAvailableSendWindow());
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
}
AckNPackets(1);
expected_send_window += kDefaultTCPMSS;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
}
TEST_F(TcpCubicBytesSenderTest, SlowStartBurstPacketLossPRR) {
sender_->SetNumEmulatedConnections(1);
// Test based on the second example in RFC6937, though we also implement
// forward acknowledgements, so the first two incoming acks will trigger
// PRR immediately.
// Ack 20 packets in 10 acks to raise the CWND to 30.
const int kNumberOfAcks = 10;
for (int i = 0; i < kNumberOfAcks; ++i) {
// Send our full send window.
SendAvailableSendWindow();
AckNPackets(2);
}
SendAvailableSendWindow();
QuicByteCount expected_send_window =
kDefaultWindowTCP + (kDefaultTCPMSS * 2 * kNumberOfAcks);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Lose one more than the congestion window reduction, so that after loss,
// bytes_in_flight is lesser than the congestion window.
size_t send_window_after_loss = kRenoBeta * expected_send_window;
size_t num_packets_to_lose =
(expected_send_window - send_window_after_loss) / kDefaultTCPMSS + 1;
LoseNPackets(num_packets_to_lose);
// Immediately after the loss, ensure at least one packet can be sent.
// Losses without subsequent acks can occur with timer based loss detection.
EXPECT_TRUE(sender_->TimeUntilSend(clock_.Now(), bytes_in_flight_,
HAS_RETRANSMITTABLE_DATA).IsZero());
AckNPackets(1);
// We should now have fallen out of slow start with a reduced window.
expected_send_window *= kRenoBeta;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Only 2 packets should be allowed to be sent, per PRR-SSRB.
EXPECT_EQ(2, SendAvailableSendWindow());
// Ack the next packet, which triggers another loss.
LoseNPackets(1);
AckNPackets(1);
// Send 2 packets to simulate PRR-SSRB.
EXPECT_EQ(2, SendAvailableSendWindow());
// Ack the next packet, which triggers another loss.
LoseNPackets(1);
AckNPackets(1);
// Send 2 packets to simulate PRR-SSRB.
EXPECT_EQ(2, SendAvailableSendWindow());
// Exit recovery and return to sending at the new rate.
for (int i = 0; i < kNumberOfAcks; ++i) {
AckNPackets(1);
EXPECT_EQ(1, SendAvailableSendWindow());
}
}
TEST_F(TcpCubicBytesSenderTest, RTOCongestionWindow) {
EXPECT_EQ(kDefaultWindowTCP, sender_->GetCongestionWindow());
// Expect the window to decrease to the minimum once the RTO fires and slow
// start threshold to be set to 1/2 of the CWND.
sender_->OnRetransmissionTimeout(true);
EXPECT_EQ(2 * kDefaultTCPMSS, sender_->GetCongestionWindow());
EXPECT_EQ(5u * kDefaultTCPMSS, sender_->GetSlowStartThreshold());
}
TEST_F(TcpCubicBytesSenderTest, RTOCongestionWindowNoRetransmission) {
EXPECT_EQ(kDefaultWindowTCP, sender_->GetCongestionWindow());
// Expect the window to remain unchanged if the RTO fires but no packets are
// retransmitted.
sender_->OnRetransmissionTimeout(false);
EXPECT_EQ(kDefaultWindowTCP, sender_->GetCongestionWindow());
}
TEST_F(TcpCubicBytesSenderTest, RetransmissionDelay) {
const int64 kRttMs = 10;
const int64 kDeviationMs = 3;
EXPECT_EQ(QuicTime::Delta::Zero(), sender_->RetransmissionDelay());
sender_->rtt_stats_.UpdateRtt(QuicTime::Delta::FromMilliseconds(kRttMs),
QuicTime::Delta::Zero(), clock_.Now());
// Initial value is to set the median deviation to half of the initial rtt,
// the median in then multiplied by a factor of 4 and finally the smoothed rtt
// is added which is the initial rtt.
QuicTime::Delta expected_delay =
QuicTime::Delta::FromMilliseconds(kRttMs + kRttMs / 2 * 4);
EXPECT_EQ(expected_delay, sender_->RetransmissionDelay());
for (int i = 0; i < 100; ++i) {
// Run to make sure that we converge.
sender_->rtt_stats_.UpdateRtt(
QuicTime::Delta::FromMilliseconds(kRttMs + kDeviationMs),
QuicTime::Delta::Zero(), clock_.Now());
sender_->rtt_stats_.UpdateRtt(
QuicTime::Delta::FromMilliseconds(kRttMs - kDeviationMs),
QuicTime::Delta::Zero(), clock_.Now());
}
expected_delay = QuicTime::Delta::FromMilliseconds(kRttMs + kDeviationMs * 4);
EXPECT_NEAR(kRttMs, sender_->rtt_stats_.smoothed_rtt().ToMilliseconds(), 1);
EXPECT_NEAR(expected_delay.ToMilliseconds(),
sender_->RetransmissionDelay().ToMilliseconds(), 1);
EXPECT_EQ(
static_cast<int64>(sender_->GetCongestionWindow() * kNumMicrosPerSecond /
sender_->rtt_stats_.smoothed_rtt().ToMicroseconds()),
sender_->BandwidthEstimate().ToBytesPerSecond());
}
TEST_F(TcpCubicBytesSenderTest, MultipleLossesInOneWindow) {
SendAvailableSendWindow();
const QuicByteCount initial_window = sender_->GetCongestionWindow();
LosePacket(acked_sequence_number_ + 1);
const QuicByteCount post_loss_window = sender_->GetCongestionWindow();
EXPECT_GT(initial_window, post_loss_window);
LosePacket(acked_sequence_number_ + 3);
EXPECT_EQ(post_loss_window, sender_->GetCongestionWindow());
LosePacket(sequence_number_ - 1);
EXPECT_EQ(post_loss_window, sender_->GetCongestionWindow());
// Lose a later packet and ensure the window decreases.
LosePacket(sequence_number_);
EXPECT_GT(post_loss_window, sender_->GetCongestionWindow());
}
TEST_F(TcpCubicBytesSenderTest, DontTrackAckPackets) {
// Send a packet with no retransmittable data, and ensure it's not tracked.
EXPECT_FALSE(sender_->OnPacketSent(clock_.Now(), bytes_in_flight_,
sequence_number_++, kDefaultTCPMSS,
NO_RETRANSMITTABLE_DATA));
// Send a data packet with retransmittable data, and ensure it is tracked.
EXPECT_TRUE(sender_->OnPacketSent(clock_.Now(), bytes_in_flight_,
sequence_number_++, kDefaultTCPMSS,
HAS_RETRANSMITTABLE_DATA));
}
TEST_F(TcpCubicBytesSenderTest, ConfigureMaxInitialWindow) {
QuicConfig config;
// Verify that kCOPT: kIW10 forces the congestion window to the default of 10.
QuicTagVector options;
options.push_back(kIW10);
QuicConfigPeer::SetReceivedConnectionOptions(&config, options);
sender_->SetFromConfig(config, Perspective::IS_SERVER);
EXPECT_EQ(10u * kDefaultTCPMSS, sender_->GetCongestionWindow());
}
TEST_F(TcpCubicBytesSenderTest, 2ConnectionCongestionAvoidanceAtEndOfRecovery) {
sender_->SetNumEmulatedConnections(2);
// Ack 10 packets in 5 acks to raise the CWND to 20.
const int kNumberOfAcks = 5;
for (int i = 0; i < kNumberOfAcks; ++i) {
// Send our full send window.
SendAvailableSendWindow();
AckNPackets(2);
}
SendAvailableSendWindow();
QuicByteCount expected_send_window =
kDefaultWindowTCP + (kDefaultTCPMSS * 2 * kNumberOfAcks);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
LoseNPackets(1);
// We should now have fallen out of slow start with a reduced window.
expected_send_window = expected_send_window * sender_->GetRenoBeta();
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// No congestion window growth should occur in recovery phase, i.e., until the
// currently outstanding 20 packets are acked.
for (int i = 0; i < 10; ++i) {
// Send our full send window.
SendAvailableSendWindow();
EXPECT_TRUE(sender_->InRecovery());
AckNPackets(2);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
}
EXPECT_FALSE(sender_->InRecovery());
// Out of recovery now. Congestion window should not grow for half an RTT.
size_t packets_in_send_window = expected_send_window / kDefaultTCPMSS;
SendAvailableSendWindow();
AckNPackets(packets_in_send_window / 2 - 2);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Next ack should increase congestion window by 1MSS.
SendAvailableSendWindow();
AckNPackets(2);
expected_send_window += kDefaultTCPMSS;
packets_in_send_window += 1;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Congestion window should remain steady again for half an RTT.
SendAvailableSendWindow();
AckNPackets(packets_in_send_window / 2 - 1);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// Next ack should cause congestion window to grow by 1MSS.
SendAvailableSendWindow();
AckNPackets(2);
expected_send_window += kDefaultTCPMSS;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
}
TEST_F(TcpCubicBytesSenderTest, 1ConnectionCongestionAvoidanceAtEndOfRecovery) {
sender_->SetNumEmulatedConnections(1);
// Ack 10 packets in 5 acks to raise the CWND to 20.
const int kNumberOfAcks = 5;
for (int i = 0; i < kNumberOfAcks; ++i) {
// Send our full send window.
SendAvailableSendWindow();
AckNPackets(2);
}
SendAvailableSendWindow();
QuicByteCount expected_send_window =
kDefaultWindowTCP + (kDefaultTCPMSS * 2 * kNumberOfAcks);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
LoseNPackets(1);
// We should now have fallen out of slow start with a reduced window.
expected_send_window *= kRenoBeta;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
// No congestion window growth should occur in recovery phase, i.e., until the
// currently outstanding 20 packets are acked.
for (int i = 0; i < 10; ++i) {
// Send our full send window.
SendAvailableSendWindow();
EXPECT_TRUE(sender_->InRecovery());
AckNPackets(2);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
}
EXPECT_FALSE(sender_->InRecovery());
// Out of recovery now. Congestion window should not grow during RTT.
for (uint64 i = 0; i < expected_send_window / kDefaultTCPMSS - 2; i += 2) {
// Send our full send window.
SendAvailableSendWindow();
AckNPackets(2);
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
}
// Next ack should cause congestion window to grow by 1MSS.
SendAvailableSendWindow();
AckNPackets(2);
expected_send_window += kDefaultTCPMSS;
EXPECT_EQ(expected_send_window, sender_->GetCongestionWindow());
}
TEST_F(TcpCubicBytesSenderTest, BandwidthResumption) {
// Test that when provided with CachedNetworkParameters and opted in to the
// bandwidth resumption experiment, that the TcpCubicSender sets initial CWND
// appropriately.
// Set some common values.
CachedNetworkParameters cached_network_params;
const QuicPacketCount kNumberOfPackets = 123;
const int kBandwidthEstimateBytesPerSecond =
kNumberOfPackets * kDefaultTCPMSS;
cached_network_params.set_bandwidth_estimate_bytes_per_second(
kBandwidthEstimateBytesPerSecond);
cached_network_params.set_min_rtt_ms(1000);
// Ensure that an old estimate is not used for bandwidth resumption.
cached_network_params.set_timestamp(clock_.WallNow().ToUNIXSeconds() -
(kNumSecondsPerHour + 1));
EXPECT_FALSE(sender_->ResumeConnectionState(cached_network_params, false));
EXPECT_EQ(10u * kDefaultTCPMSS, sender_->GetCongestionWindow());
// If the estimate is new enough, make sure it is used.
cached_network_params.set_timestamp(clock_.WallNow().ToUNIXSeconds() -
(kNumSecondsPerHour - 1));
EXPECT_TRUE(sender_->ResumeConnectionState(cached_network_params, false));
EXPECT_EQ(kNumberOfPackets * kDefaultTCPMSS, sender_->GetCongestionWindow());
// Resumed CWND is limited to be in a sensible range.
cached_network_params.set_bandwidth_estimate_bytes_per_second(
(kMaxTcpCongestionWindow + 1) * kDefaultTCPMSS);
EXPECT_TRUE(sender_->ResumeConnectionState(cached_network_params, false));
EXPECT_EQ(kMaxTcpCongestionWindow * kDefaultTCPMSS,
sender_->GetCongestionWindow());
cached_network_params.set_bandwidth_estimate_bytes_per_second(
(kMinCongestionWindowForBandwidthResumption - 1) * kDefaultTCPMSS);
EXPECT_TRUE(sender_->ResumeConnectionState(cached_network_params, false));
EXPECT_EQ(kMinCongestionWindowForBandwidthResumption * kDefaultTCPMSS,
sender_->GetCongestionWindow());
// Resume to the max value.
cached_network_params.set_max_bandwidth_estimate_bytes_per_second(
(kMinCongestionWindowForBandwidthResumption + 10) * kDefaultTCPMSS);
EXPECT_TRUE(sender_->ResumeConnectionState(cached_network_params, true));
EXPECT_EQ((kMinCongestionWindowForBandwidthResumption + 10) * kDefaultTCPMSS,
sender_->GetCongestionWindow());
}
TEST_F(TcpCubicBytesSenderTest, PaceBelowCWND) {
QuicConfig config;
// Verify that kCOPT: kMIN4 forces the min CWND to 1 packet, but allows up
// to 4 to be sent.
QuicTagVector options;
options.push_back(kMIN4);
QuicConfigPeer::SetReceivedConnectionOptions(&config, options);
sender_->SetFromConfig(config, Perspective::IS_SERVER);
sender_->OnRetransmissionTimeout(true);
EXPECT_EQ(kDefaultTCPMSS, sender_->GetCongestionWindow());
EXPECT_TRUE(sender_->TimeUntilSend(QuicTime::Zero(), kDefaultTCPMSS,
HAS_RETRANSMITTABLE_DATA).IsZero());
EXPECT_TRUE(sender_->TimeUntilSend(QuicTime::Zero(), 2 * kDefaultTCPMSS,
HAS_RETRANSMITTABLE_DATA).IsZero());
EXPECT_TRUE(sender_->TimeUntilSend(QuicTime::Zero(), 3 * kDefaultTCPMSS,
HAS_RETRANSMITTABLE_DATA).IsZero());
EXPECT_FALSE(sender_->TimeUntilSend(QuicTime::Zero(), 4 * kDefaultTCPMSS,
HAS_RETRANSMITTABLE_DATA).IsZero());
}
} // namespace test
} // namespace net