Can on-demand origin-pull be supported in srs edge mode, edge standby. #419
Comments
|
Can you provide real examples? You can omit key client information.
|
|
Explain the difficulties you are currently facing, without any assumptions about why the user needs this feature. It should be something like this: describe the problem you are currently facing, what the customer needs, what you need to provide, where you are stuck, and how you have managed to work around it. It would be convenient if the SRS (Software Requirements Specification) could support certain aspects. Something along these lines.
|
|
This can be called: edge standby mode, waiting for the origin server to request streaming before starting to stream.
|
|
The previous mode was RTSP, with each server being a source station. Now, after switching to the SRS cluster, the stream is pushed to the edge and delivered directly to the source station; unlike the streams played at the edge, which are sourced on demand.
|
|
Waiting for implementation. This feature is too basic. I can't believe it hasn't been done, have they not considered the bandwidth? The situation I encountered is similar to the example above. I have to provide access to hundreds of cameras for a certain boss to view at any time, but most of the time, he doesn't even look at them. Occasionally, he might only view one of them. Although it is said that it takes a long time to raise an army but only a moment to use it, sending several hundred video streams from his internal network to my public network server is completely unsustainable. If it is pushed on-demand, it would be more reasonable to temporarily push the stream of the camera he wants to view.
|
|
Waiting for PR.
|
|
There is one implementation method where, during standby, after the edge connects to the origin server, it only sends ping packets to maintain the connection, or sends control message packets, or empty packets, in short, it keeps the data volume from the edge to the origin server very small, thus supporting a very large number of connections. When there is a need for source playback, the origin server sends control messages to the edge, and the edge then delivers the stream.
|
|
Looking back at this feature, it is a requirement for monitoring scenes. Generally, there will be additional signaling. After the signaling interaction for watching or adjusting the camera position is completed, the stream will be sent to the media server. This requirement is merged into #1500 gb18181 for support.
|
|
Please use GB28181 #1500 |
winlinvip
changed the title
srs edge mode, can it push to the origin server only when the origin server needs the stream, to reduce bandwidth waste.
Current srs: fetch stream from origin server (during playback) or forward stream to origin server (during publishing).
On-demand fetch srs: fetch stream from origin server (during playback) or forward stream to origin server (during origin server playback).
【Leader】Cute Leiwangbao 2015/6/5 16:45:16
It's just that the configuration in SRS is hardcoded... like static routing.
16:47:54
【Leader】Winlin 2015/6/5 16:47:54
Add a standby function and report a bug.
【Grandmaster】Hunan-Front Desk-Xiaobai
Let me give a simple example... There is a unit with 100 surveillance streams, and they have a command center that only plays 16 streams in real-time.
【Grandmaster】BK 2015/6/5 16:49:20
This example is very good.
Real-life scenario:
The customer's factory has 1000 cameras, which are published to the edge servers in the factory for real-time recording.
The source server publicly plays the streams, providing services such as factory visits and 3D factory displays. The source server serves not only one factory but also 10 factories, with a total of 10,000 streams. In this case, the source server does not want all 1000 streams to be fetched from the origin server. It would be better to fetch them from the origin server only when there is a play request, in order to reduce the load on the origin server.
TRANS_BY_GPT3The text was updated successfully, but these errors were encountered: