Distributed Scene Graph Asset De-duplication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud computing faces challenges with high data bandwidth requirements for transmitting and storing assets, duplication of data leading to inefficient storage and processing, and the inability to easily upgrade standard assets.
Innovation Solution
A distributed file format that uses unique hash representations to de-duplicate assets and allows referencing from scene graph files, reducing data size and storage needs, and enables efficient traversal and rendering by storing assets close to compute nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If all assets are transmitted and stored in the cloud for computation, then computation can be performed remotely, but data bandwidth requirements become prohibitively high and storage costs increase
Solution Approach 1:
The patent segments assets into two categories: standard assets stored in a cloud-based library and custom assets stored locally. Scene graphs reference standard assets by ID rather than containing full copies, reducing transmitted data while maintaining remote computation capability. This segmentation allows the system to optimize bandwidth usage by only transmitting scene graph structures and custom asset references.
Solution Approach 2:
The patent creates lightweight copies of asset references in the form of scene graphs that contain only metadata and references to the actual asset data. These scene graphs are transmitted to cloud compute nodes, which then retrieve the actual asset data from the cloud asset library during rendering, significantly reducing the bandwidth required for asset transmission.
2Productivity
If all asset data is stored in the cloud, then data can be reused across multiple computations, but data duplication occurs and storage costs increase
Solution Approach 1:
The patent extracts the actual asset data from individual scene graphs and stores it centrally in a cloud-based asset library. Scene graphs contain only references (asset IDs) to the actual data, not the data itself. This extraction eliminates duplication while maintaining reuse efficiency, as the same asset data can be referenced by multiple scene graphs without being stored multiple times.
3Productivity
If standard assets are uploaded for prior computation, then computation can proceed, but upgrading those assets requires re-uploading and re-computation
Solution Approach 1:
The patent introduces an intermediary layer (the cloud asset library) between the scene graphs and the actual asset data. Scene graphs reference assets by ID in the library rather than containing embedded copies. This intermediary allows asset upgrades to be performed by simply replacing the asset data in the library, while all scene graphs automatically reference the updated version, enabling seamless asset updates without re-computation.
Data Source
AI summary
A method, apparatus, system, and computer program product provide the ability to utilize a distributed scene graph. A database of hashes and corresponding location identifiers (for asset locations) is maintained. An export of a scene (including scene assets) is initiated and hash values for the scene assets are computed. If the hash value is not in the database, it is added to the database. If the hash value is in the database, the corresponding location identifier is used to identify the location of the scene asset. The hash value and corresponding location identifier are stored in a scene graph file that is traversed. During the traversal, for each hash value in the scene graph file, the scene asset is retrieved from the location identified in the scene graph file and processed.