Nonessential Task Signaling in NBMP Cloud Workflows
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The MPEG Network Based Media Processing (NBMP) standard lacks distinction between essential and nonessential tasks, fails to support media streams with variable durations, and does not allow exclusion of function instances from workflow descriptions, limiting its ability to efficiently process media in cloud environments.
Innovation Solution
Implementing a signaling mechanism to identify nonessential inputs, outputs, and tasks, extending the NBMP standard to support independent processing of media segments with variable durations, and enabling exclusion of function instances from workflow descriptions to optimize media processing workflows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the NBMP standard processes all tasks in a workflow as essential, then workflow completeness is maintained, but processing efficiency decreases due to unnecessary execution of nonessential tasks
Solution Approach 1:
The workflow is segmented into essential and nonessential tasks through syntax elements. The workflow manager can selectively execute only essential tasks (those without non_essential_flag set) while skipping nonessential tasks, thereby improving processing efficiency without compromising the completion of critical workflow functions.
Solution Approach 2:
The system performs partial action by executing only the necessary subset of tasks (essential tasks) rather than all tasks in the workflow. This is achieved through the non_essential_flag syntax element that allows the workflow manager to identify and skip nonessential tasks, reducing unnecessary processing while maintaining workflow effectiveness.
2Adaptability or versatility
If the NBMP standard uses fixed duration segments, then processing simplicity is maintained, but adaptability to variable duration media streams is lost
Solution Approach 1:
The segment processing system transitions from static fixed-duration segments to dynamic variable-duration segments. The step_descriptor syntax element enables the workflow to specify different duration values for different segments, allowing the system to adapt to variable duration media streams while maintaining a relatively simple processing framework.
3Ease of operation
If all function instances are included in workflow descriptions, then comprehensive processing is ensured, but workflow management complexity increases
Solution Approach 1:
Nonessential function instances are extracted and identified using the non_essential_flag syntax element in the workflow description. This allows the workflow manager to exclude these function instances from execution, simplifying workflow management by removing unnecessary components while maintaining the essential processing functions.
4Productivity
If multidimensional segmentation is implemented, then processing parallelism is increased, but system complexity increases
Solution Approach 1:
The system extends segmentation from a single temporal dimension to multiple dimensions including spatial dimensions (width, height) and depth. This multidimensional segmentation enables increased processing parallelism by allowing independent processing of segments across multiple dimensions simultaneously, while the step_descriptor provides a unified framework to manage this complexity.
Data Source
AI summary
Systems and methods for managing a Network Based Media Processing (NBMP) workflow are provided. A method includes a method performed by at least one processor is provided. The method includes: deriving the NBMP workflow; obtaining at least one first syntax element indicating that at least one task included in the NBMP workflow, at least one input received by the at least one task, or at least one output generated by the at least one task, is nonessential; determining a plurality of essential tasks based on the at least one first syntax element; and assigning the plurality of essential tasks to at least one from among a media sink, a media source, and a media processing entity included in the NBMP workflow.