若要更加全面展示多流共存场景的动态,需更新了一下仿真函数,模拟新流侵入和退出场景:
void simulate_one_phase(int i)
{
double rtt;
bbr_update_sending_bw(&f1);
bbr_update_sending_bw(&f2);
bbr_update_sending_bw(&f3);
bbr_update_sending_bw(&f4);
printf("t= %04d sending: f1: %.3f f2: %.3f f3: %.3f f4: %.3f\n",
t, f1.sending_bw, f2.sending_bw, f3.sending_bw, f4.sending_bw);
if (i < 1000) {
rtt = (f1.inflt + f2.inflt + f3.inflt) / C;
f1.receive_bw = C * f1.inflt / (f1.inflt + f2.inflt + f3.inflt);
f2.receive_bw = C * f2.inflt / (f1.inflt + f2.inflt + f3.inflt);
f3.receive_bw = C * f3.inflt / (f1.inflt + f2.inflt + f3.inflt);
f4.receive_bw = 1;
f4.max_bw = 1;
f4.inflt = 1;
if (i == 999) {
f4.max_bw = 0.9 * C;
f4.inflt = 0.9 * C * RTPROP;
}
} else if (i > 1000 && i < 2000) {
rtt = (f1.inflt + f2.inflt + f3.inflt + f4.inflt) / C;
f1.receive_bw = C * f1.inflt / (f1.inflt + f2.inflt + f3.inflt + f4.inflt);
f2.receive_bw = C * f2.inflt / (f1.inflt + f2.inflt + f3.inflt + f4.inflt);
f3.receive_bw = C * f3.inflt / (f1.inflt + f2.inflt + f3.inflt + f4.inflt);
f4.receive_bw = C * f4.inflt / (f1.inflt + f2.inflt + f3.inflt + f4.inflt);
} else {
rtt = (f1.inflt + f2.inflt) / C;
f1.receive_bw = C * f1.inflt / (f1.inflt + f2.inflt);
f2.receive_bw = C * f2.inflt / (f1.inflt + f2.inflt);
f3.receive_bw = 0;
f4.receive_bw = 0;
f3.max_bw = 0;
f4.max_bw = 0;
f3.inflt = 0;
f4.inflt = 0;
}
printf("t= %04d receive: f1: %.3f f2: %.3f f3: %.3f f4: %.3f\n",
t, f1.receive_bw, f2.receive_bw, f3.receive_bw, f4.receive_bw);
bbr_update_maxbw_minrtt(&f1, rtt);
bbr_update_maxbw_minrtt(&f2, rtt);
bbr_update_maxbw_minrtt(&f3, rtt);
bbr_update_maxbw_minrtt(&f4, rtt);
printf("t= %04d max_bw: f1: %.3f f2: %.3f f3: %.3f f4: %.3f\n",
t, f1.max_bw, f2.max_bw, f3.max_bw, f4.max_bw);
printf("t= %04d inflt: f1: %.3f f2: %.3f f3: %.3f f4: %.3f\n",
t, f1.inflt, f2.inflt, f3.inflt, f4.inflt);
printf("t= %04d min_rtt: f1: %.3f f2: %.3f f3: %.3f f4: %.3f\n",
t, f1.min_rtt, f2.min_rtt, f3.min_rtt, f4.min_rtt);
printf("t= %04d pacing_gain: f1: %.3f f2: %.3f f3: %.3f f4: %.3f\n\n",
t, f1.pacing_gain, f2.pacing_gain, f3.pacing_gain, f4.pacing_gain);
t++;
bw_filter_index = (bw_filter_index + 1) % BW_FILTER_LEN;
}
它的意思是,在前 1/3 时间,3 流共存,此后侵入 1 条新流,维持 1/3 时间,最后 1/3 时间退出 2 条流,维持 2 条共存。C = 100Mbps,R = 2 时,delivery rate,pacing rate,inflt 的仿真结果如下:
与 Linux kernel 的 tcp_bbr.c 相比,bbr-sim.c 简单又清晰,其实它本来就简单清晰,多数复杂性都是与 linux tcp 处理框架之间的接口引入,比如 delivery rate 的采集,srtt 采集。而这一切则可通过抽象而避免。
简单来讲,一个方程即可描述 bbr,设 x 为 inflt,C,R 分别为瓶颈带宽和传播时延:
d x d t = C ⋅ R − x \dfrac{dx}{dt}=C\cdot R-x dtdx=C⋅R−x
剩下的就是实现细节。把排队时延抽象到全局 inflt / C,而不是具象到 buffer,这样就不再需要将 rtt 分别排队时延和传播时延两部分,也不再需要分别处理。
浙江温州皮鞋湿,下雨进水不会胖。