kwaivideoteam/las

LAS (Live Adaptive Streaming), an adaptive bitrate streaming technique that enables high quality and low-latency live streaming of media content over HTTP, WebRTC, and so on.

las's Issues

直接跑srs不成功

部署了srs然后把manifest.json的url改了报202-source buffer error错误，能不能把转码视频上传下，是不是codec有问题？
ffmpeg.exe -re -i test.mp4 -vcodec libx264 -vf scale=1280:720 -b:v 2000k -x264-params "no-scenecut=1:threads=8:keyint=90" -acodec aac -f flv rtmp://192.168.6.62:1935/live/out0 -vcodec libx264 -vf scale=960:540 -b:v 1000k -x264-params "no-scenecut=1:threads=8:keyint=90" -acodec aac -f flv rtmp://192.168.6.62:1935/live/out1 -vcodec libx264 -vf scale=640:360 -b:v 500k -x264-params "no-scenecut=1:threads=8:keyint=90" -acodec aac -f flv rtmp://192.168.6.62:1935/live/out2

SRS 转码配置失败

SRS: 4.0.29(Leo)
Commit: 53326d8

部署配置：

listen              1935;
max_connections     1000;
daemon              off;
srs_log_tank        console;
srs_log_level       trace;

http_server {
    enabled         on;
    listen          8080;
    dir             ./objs/nginx/html;
}

http_api {
    enabled         on;
    listen          1985;
}

vhost __defaultVhost__ {

    tcp_nodelay     on
    min_latency     on;
    
    play {
        gop_cache       on;
        queue_length    10;
        mw_latency      100;
    }

    publish {
        mr          off;
    }
    
    http_remux {
        enabled     on;
        mount       [vhost]/[app]/[stream].flv;
    }

    atc             on;
    transcode {
        enabled     on;
        ffmpeg      /usr/bin/ffmpeg;
        
        engine low {
            enabled         on;
            vcodec          libx264;
            vbitrate        600;
            vwidth          360;
            vheight         640;
            vthreads        12;
            vprofile        main;
            vpreset         veryfast;
            vparams {
                -x264-params "no-scenecut=1:keyint=90";
            }
            acodec          copy;
            output          rtmp://127.0.0.1:[port]/[app]?vhost=[vhost]/[stream]_[engine];
        }

        engine medium {
            enabled         on;
            vcodec          libx264;
            vbitrate        1200;
            vwidth          540;
            vheight         960;
            vthreads        12;
            vprofile        main;
            vpreset         veryfast;
            vparams {
                -x264-params "no-scenecut=1:keyint=90";
            }
            acodec          copy;
            output          rtmp://127.0.0.1:[port]/[app]?vhost=[vhost]/[stream]_[engine];
        }

        engine high {
            enabled         on;
            vcodec          libx264;
            vbitrate        1600;
            vwidth          720;
            vheight         1280;
            vthreads        12;
            vprofile        main;
            vpreset         veryfast;
            vparams {
                -x264-params "no-scenecut=1:keyint=90";
            }
            acodec          copy;
            output          rtmp://127.0.0.1:[port]/[app]?vhost=[vhost]/[stream]_[engine];
        }
    }
}

推流：OBS 26.0.2 (MacOS)

错误讯息：

[2020-11-24 17:54:17.771][Warn][31543][233][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.771][Warn][31543][233][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.771][Warn][31543][234][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.771][Warn][31543][234][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.771][Warn][31543][234][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.771][Warn][31543][234][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.771][Warn][31543][234][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.772][Trace][31543][238] HTTP client ip=61.220.72.232, request=0, to=15000ms
[2020-11-24 17:54:17.772][Error][31543][238][11] serve error code=3009 : parse message : parse 517B, nparsed=0, err=16/HPE_INVALID_METHOD invalid HTTP method
thread [31543][238]: parse_message() [src/service/srs_service_http_conn.cpp:95][errno=11]
thread [31543][238]: parse_message_imp() [src/service/srs_service_http_conn.cpp:127][errno=11](Resource temporarily unavailable)
[2020-11-24 17:54:17.782][Trace][31543][239] HTTP client ip=61.220.72.232, request=0, to=15000ms
[2020-11-24 17:54:17.782][Error][31543][239][11] serve error code=3009 : parse message : parse 517B, nparsed=0, err=16/HPE_INVALID_METHOD invalid HTTP method
thread [31543][239]: parse_message() [src/service/srs_service_http_conn.cpp:95][errno=11]
thread [31543][239]: parse_message_imp() [src/service/srs_service_http_conn.cpp:127][errno=11](Resource temporarily unavailable)
[2020-11-24 17:54:17.901][Warn][31543][232][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.901][Warn][31543][232][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.901][Warn][31543][232][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.901][Warn][31543][232][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.901][Warn][31543][232][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.905][Warn][31543][233][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.905][Warn][31543][233][11] aac ignore type=1 for no sequence header
[2020-11-24 17:54:17.905][Warn][31543][233][11] aac ignore type=1 for no sequence header

我用原始 SRS4.0 release 部署配置是可行的，也不会有这些错误讯息，是什么原因呢？

defaultLiveDelay指端到端延迟还是播放缓冲区？

LAS have support edge server like http-flv in SRS

I want to scaling my project using LAS. But i wonder how can i using edge server like http-flv

_getPredictedBuffer和_getBufferSpeed

las/client/las.js/src/abr/abr-algorithm-simple.ts

Line 243 in 33335b2

private _getPredictedBuffer(buffered: number) {

las/client/las.js/src/abr/abr-algorithm-simple.ts

Line 248 in 33335b2

private _getBufferSpeed(buffered: number) {

你好，这两个function看起来不太理解，能不能稍微解释下呢？

支持WebSocket-FLV么？

最高最低码率设置一样的话是不是可以当作一个flv.js

带宽预测

docs/LAS1.0.pfg的 6.3.2网络状态收集有一段话
在网络状态收集层面，采用固定时间采点 的模式，即每隔一个固定时间𝑇(𝑚𝑠)，统计该时间段实际下载的数据量 𝑆(𝐵𝑦𝑡𝑒𝑠)，从而得到一个带宽采样点𝐵(𝑘𝑏𝑝𝑠) = 𝑆 ∗ 8/𝑇。典型的，𝑇 = 500𝑚𝑠。 基于这些带宽的采样点，通过滤波和预测算法，估计网络的真实带宽，作为 码率调整的依据。 注:带宽估计的参考实现逻辑，将在参考代码给出
然后我查看了ijklas.c的OnStatisticInfo方法，RateAdaptionState的bw_c也是根据统计固定时间间隔的下载数据来估算的，也就是说只有在卡顿状态下，该值近似与带宽，而网络状态良好时，该值与流的码率大致相同，不能反应实际带宽。

请问这里是失误还是故意没有给出基于卡尔曼滤波的代码实现吗？

大烂尾项目

和LHLS比较怎样？

感觉不如LHLS简洁，CDN厂商要支持las，改动太大。