Autonomous Compute Storage Engine Offloads CPU via Read Signature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computational storage systems face limitations due to their reliance on CPU-centric paradigms, requiring modification of existing software, limited developer expertise, compatibility issues across different storage devices, and varying deployment configurations, which hinder the widespread adoption of computational storage capabilities.
Innovation Solution
An Information Handling System (IHS) with a storage device chassis that includes a processing system and memory, capable of autonomously executing compute operations by receiving read instructions, identifying data, and executing associated compute applications based on autonomous compute signatures, thereby offloading compute tasks from the CPU.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional computational storage systems use CPU-centric paradigms with storage device compute applications, then compute operations can be offloaded from the CPU, but the system requires modification of existing software, limited developer expertise, and has compatibility issues across different storage devices
Solution Approach 1:
The patent inverts the conventional approach by making the storage device the autonomous decision-maker rather than the CPU. Instead of the CPU directing compute operations through complex software paradigms, the storage device autonomously determines when and how to perform compute operations by monitoring host read instructions and data patterns, eliminating the need for CPU-centric software modifications
Solution Approach 2:
The storage device performs self-service by autonomously executing compute operations without CPU direction. The device monitors its own operational state, identifies opportunities for compute operations based on host read instructions and data characteristics, and executes appropriate compute applications independently, reducing reliance on external CPU control and complex software infrastructure
2Productivity
If storage devices execute compute applications autonomously, then processing efficiency is improved, but the device must handle multiple configurations and compatibility across different storage devices
Solution Approach 1:
The patent applies preliminary action by having the storage device pre-monitor host read instructions and data patterns before executing compute operations. The device proactively identifies suitable compute opportunities based on predetermined criteria embedded in the autonomous compute engine, allowing it to adapt to different configurations without requiring real-time complex decision-making or extensive compatibility handling
3Ease of operation
If conventional systems require CPU sequencing of compute operations, then developer control is maintained, but the system has limited developer expertise requirements and compatibility constraints
Solution Approach 1:
The autonomous compute engine performs self-service by independently monitoring host operations and determining when to execute compute applications. This eliminates the need for developers to implement complex CPU sequencing logic and reduces expertise requirements for NVMe programming constructs and storage device stack modifications
Solution Approach 2:
The patent extracts the compute operation sequencing function from the CPU and relocates it to the storage device's autonomous compute engine. This extraction simplifies the overall system by removing the need for CPU-directed sequencing complexity while maintaining developer control through the host's ability to issue read instructions
Data Source
AI summary
An autonomous compute storage device system includes a computing device and a storage device that is coupled to the computing device. The storage device receives a read instruction from a host processing system in the computing device that identifies data stored in a storage subsystem included in the storage device and, in response, performs a read operation to copy the data from the storage subsystem to a memory subsystem accessible to the storage device and provide the data to the host processing system. If the storage device determines that an autonomous compute signature matches the data that was copied to the memory subsystem during the performance of the read operation, it executes an autonomous compute application to perform compute operations that are associated with the data that was copied to the memory subsystem during the performance of the read operation and generate compute operation result(s).