Media Workload Scheduler for GPU Bottleneck Resolution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Media workloads on graphics processing units (GPUs) can exceed maximum throughput capacity, leading to bottlenecks and disruptions in data processing, such as video playback lag or interruptions, due to high GPU utilization and memory bandwidth usage.

Innovation Solution

Implementing a method to dynamically adjust media workloads by modifying features like decoding and encoding processes, disabling in-loop de-blocking, skipping certain frames, and reducing bit rate and macroblock size, while prioritizing feature disabling to minimize impact on video quality and GPU utilization, using a workload scheduler and media feature modeling database to optimize GPU and memory bandwidth usage.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If media workload is increased to fully utilize GPU time and memory bandwidth, then processing capacity is improved, but bottlenecks and disruptions occur in data processing

Engineering Contradiction:

ImproveGPU processing capacityVSAvoiddata processing stability

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system dynamically adjusts media workload parameters based on real-time GPU utilization metrics. The workload scheduler continuously monitors GPU usage and modifies encoding/decoding parameters, frame rates, and resolution settings to maintain optimal performance while preventing bottlenecks, transforming the static workload into a dynamic adaptive system.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes multiple workload parameters including encoding bitrate, resolution, frame rate, and macroblock size to optimize GPU utilization. By adjusting these parameters dynamically, the system can increase processing capacity when GPU is underutilized and reduce parameters when approaching bottleneck thresholds, resolving the contradiction between productivity and reliability.

Inventive Principle:

Principle #35Parameter changes

2Productivity

If GPU utilization is increased to maximize media processing throughput, then productivity is improved, but video playback disruptions occur

Engineering Contradiction:

Improvemedia processing throughputVSAvoidvideo playback continuity

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The workload scheduler implements a feedback mechanism that continuously monitors GPU utilization metrics and video playback performance. When disruptions are detected or GPU utilization exceeds optimal thresholds, the system automatically adjusts workload parameters to restore smooth playback, creating a closed-loop control system that balances throughput and continuity.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary adjustments to workload parameters before bottlenecks occur by monitoring GPU utilization trends. By proactively reducing encoding complexity or frame rates when utilization approaches critical levels, the system prevents playback disruptions before they occur, maintaining both high throughput and smooth video delivery.

Inventive Principle:

Principle #10Preliminary action

3Reliability

If media workload features are reduced to prevent bottlenecks, then processing stability is improved, but video quality may degrade

Engineering Contradiction:

Improveprocessing stabilityVSAvoidvideo quality

Core Design Contradiction:

ReliabilityVSManufacturing precision

Solution Approach 1:

The system applies different quality levels to different parts of the video processing pipeline based on local conditions. Instead of uniformly reducing all parameters, the workload scheduler selectively adjusts specific features like disabling in-loop de-blocking or reducing macroblock size only when and where necessary, preserving video quality in areas where GPU capacity allows while maintaining stability where bottlenecks occur.

Inventive Principle:

Principle #3Local quality

4Productivity

If workload parameters are dynamically adjusted to optimize GPU utilization, then processing efficiency is improved, but system complexity increases

Engineering Contradiction:

Improveprocessing efficiencyVSAvoidworkload scheduling complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The workload scheduler implements self-service mechanisms where the system automatically monitors its own performance metrics and adjusts parameters without external intervention. The GPU utilization monitoring and parameter adjustment logic is embedded within the media processing pipeline itself, allowing the system to self-optimize based on real-time conditions without requiring complex external control systems.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS8928678B2Media workload scheduler

Publication Date: 2015.01.06 TAHOE RES LTD

AI summary

A method and system for scheduling a media workload is disclosed herein. The method includes modeling a feature of the media workload. A GPU utilization rate and a memory bandwidth of the media workload may be determined. Additionally, the media workload may be scheduled by modifying the feature of the media workload in order to adjust the GPU utilization and the memory bandwidth.