Livestream Augmentation via Dynamic Processor Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for livestream transformation and augmentation face challenges in scalability and real-time processing, particularly due to compute constraints on streamer devices and the need for proprietary models with complex installation and deployment requirements, while also incurring unnecessary computation costs and resource inefficiencies.
Innovation Solution
A scalable and resource-efficient system that applies transformations to livestreams using a dedicated server, dynamically determining the number of processors for parallel processing based on available resources and latency budgets, minimizing idle compute time and avoiding unnecessary decoding and re-encoding, and utilizing machine learning models for real-time modifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If transformations are applied using a dedicated server with dynamic processor allocation, then scalability and resource efficiency are improved, but system complexity increases
Solution Approach 1:
The patent introduces a dedicated transformation server as an intermediary between streamer devices and streaming services. This server handles complex transformation operations using dynamically allocated processors, isolating the complexity from streamer devices while maintaining scalability through centralized resource management.
Solution Approach 2:
The system dynamically determines the number of processors to allocate based on available resources and latency budgets. This dynamic resource allocation allows the system to scale efficiently by adjusting computing power according to actual transformation needs and service level agreements.
2Speed
If parallel processing with dynamically determined processors is used, then real-time processing capability is improved, but compute resource management complexity increases
Solution Approach 1:
The system performs preliminary actions by pre-establishing latency budgets and resource allocation policies before transformations are applied. This allows parallel processing to proceed efficiently with predetermined guidance, reducing the complexity of real-time resource management while maintaining processing speed.
3Manufacturing precision
If proprietary models with complex installation requirements are deployed, then transformation quality is improved, but ease of deployment worsens
Solution Approach 1:
The patent extracts proprietary transformation models from streamer devices and concentrates them on the dedicated transformation server. This extraction allows high-quality proprietary models to be deployed centrally without burdening individual streamer devices with complex installation requirements, improving both transformation quality and ease of deployment.
4Loss of time
If transformations are applied at the streamer device, then latency is reduced, but device performance and other applications are degraded
Solution Approach 1:
The system segments the transformation processing function from the streamer device, placing it on a dedicated server. This segmentation allows the streamer device to focus on broadcasting while the server handles transformations, reducing latency through efficient parallel processing without degrading device performance or affecting other applications.
Data Source
AI summary
Systems and methods are described that include receiving, at a streaming server, a data stream; upon determining, by the streaming server, a transformation to be applied to the data stream, transmitting the data stream to a processing server and applying, using the processing server, the transformation to the data stream; generating a transformed data stream for use by a platform. Systems and methods can include applying, using a determined number of processors, the transformation to a plurality of data stream samples, each processor of the determined number of processors applying the transformation to a data stream sample and generating a transformed data stream sample. Systems and methods can apply, using a processor, the transformation to a plurality of data stream sample sequences and generate corresponding transformed data stream sample sequences. Determining the number of processors can be based on a latency budget value and/or an attribute value of the transformation.