Storage Device In-Storage Graphic Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The load on hosts increases during graphic processing operations due to the need to handle large amounts of graph data, degrading graphic processing speed and efficiency in storage systems.
Innovation Solution
A storage device with non-volatile memory that stores source and destination vertex values, allowing for in-storage graphic processing operations by reading and processing vertex values directly within the storage device, reducing the load on the host through efficient vertex search and processing operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If graphic processing operations are performed by the host, then the host can process graph data, but the load on the host increases and processing speed decreases
Solution Approach 1:
The patent extracts the graphic processing function from the host system and relocates it to the storage device. Specifically, the storage controller is enhanced to perform vertex search operations and graphic processing tasks locally, separating this computational workload from the host. This extraction resolves the contradiction by maintaining processing capability while eliminating the burden on the host system.
Solution Approach 2:
The storage controller acts as an intermediary between the host and the non-volatile memory. Instead of the host directly processing graph data from storage, the storage controller intermediates by performing preliminary vertex searches and processing operations on the graph data before presenting results to the host. This intermediary function improves processing speed while reducing host load.
2Reliability
If large amounts of graph data are transferred for processing, then complete data is available for processing, but transfer time increases and processing efficiency decreases
Solution Approach 1:
Instead of transferring and processing all graph data completely, the patent applies partial action by performing vertex search operations directly within the storage device. The storage controller processes only the specific vertex data needed for the current operation, rather than moving entire datasets. This maintains data reliability for the required operations while dramatically reducing transfer time.
Solution Approach 2:
The storage controller performs preliminary processing actions on graph data while it remains in the non-volatile memory. Vertex search operations and preliminary graphic processing are executed before data would normally be transferred to the host, preparing the data in advance and reducing the amount of work needed after transfer.
3Measurement precision
If vertex search operations are performed in the host, then accurate results are obtained, but processing time increases
Solution Approach 1:
The patent substitutes the mechanical data transfer process with a more efficient operational approach. Instead of transferring all graph data to the host for vertex search operations, the system uses the storage controller's processing capabilities to perform vertex searches directly in the storage device. This substitution maintains search accuracy while improving processing speed by eliminating data transfer bottlenecks.
Data Source
AI summary
Provided is a storage device including non-volatile memory and a storage controller, where the non-volatile memory includes a plurality of cell strings storing a plurality of source vertex values and a plurality of destination vertex values, and the storage controller reads at least a first destination vertex value corresponding to a first source vertex value from the non-volatile memory, and performs a graphic processing operation based on the read first destination vertex value. A first cell string includes a plurality of first memory cells storing a first source vertex value and connected to word lines of a first group of wordlines, respectively, and at least one second memory cell storing the first destination vertex value and connected to at least one word line of a second group of wordlines.