Live Streaming Slicing for Latency Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current HTTP Live Streaming (HLS) protocols face challenges in reducing latency and enhancing user experience due to delays in live stream delivery, particularly during cold streaming starts, and inefficiencies in server and storage resource utilization.
Innovation Solution
The method involves an HLS source station server acquiring live streaming, slicing it into index and sliced files, and storing these files in an object storage server, allowing for efficient retrieval and playback, with two modes: actively triggering slices to eliminate initial delays and passively triggering slices to conserve resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If the HLS protocol pulls stream from edge node to source station server and locally converts to generate M3U8 and TS files, then the live stream can be delivered to users, but the latency increases due to multiple levels of returning to source station
Solution Approach 1:
The source station server proactively pushes sliced stream files to the object storage server in advance, before users request the stream. This eliminates the need for edge nodes to repeatedly request and wait for stream files from source station, significantly reducing latency. The slicing operation is performed once at the source station and results are cached in object storage for rapid delivery.
Solution Approach 2:
An object storage server is introduced as an intermediary between the source station server and edge nodes. The object storage server stores sliced stream files and serves them directly to edge nodes, eliminating the need for edge nodes to communicate directly with the source station server for each stream request. This reduces the number of communication hops and latency.
2Loss of time
If the source station server actively pushes sliced files to object storage server, then latency is reduced and user experience improves, but server and storage resource consumption increases
Solution Approach 1:
The source station server performs slicing and pushes files to object storage in advance, preparing content before it is requested. This eliminates first-screen delay when users access the stream, as files are already available in object storage. The trade-off of increased resource usage is acceptable because it improves user experience and reduces the need for repeated processing.
Solution Approach 2:
The sliced stream files are copied from the source station server to the object storage server. Instead of having edge nodes request and download files repeatedly from the source station, a copy is created and stored in object storage, which can serve multiple requests simultaneously. This distributes the storage burden and enables faster delivery.
3Reliability
If edge nodes request stream files from source station server for each user request, then fresh stream content is delivered, but the response time increases due to cold starts and multiple levels of returning
Solution Approach 1:
The source station server pre-slices the stream and pushes files to object storage in advance, so when users request the stream, the files are already prepared and available for immediate delivery. This eliminates cold start delays while maintaining content freshness through continuous slicing of new stream data.
Solution Approach 2:
The object storage server acts as an intermediary that caches sliced stream files, allowing edge nodes to retrieve files quickly without directly contacting the source station server for each request. This maintains content freshness while dramatically improving response speed by reducing communication latency and the number of request-hold-wait cycles.
Data Source
AI summary
Embodiments of the present disclosure relate to a method, apparatus, and system for slicing live streaming. A method may include: acquiring the live streaming from a live source station server; slicing the live streaming to generate an index file and sliced files of the live streaming; and sending the index file and the sliced files of the live streaming to an object storage server, to cause the object storage server to store the index file and the sliced files of the live streaming.