SSD Hardware Accelerator for Database Sorting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The performance of sorting algorithms in solid state storage drives (SSDs) is limited by the lack of computational resources and RAM in CPUs, leading to increased power consumption and communication interface utilization during external sorting processes, which affects the efficiency of data retrieval and processing.
Innovation Solution
An embedded reconfigurable SSD hardware accelerator offloads sorting operations from host RAM and CPU resources, utilizing an upgradable sorting algorithm to generate and store sorted data subsets within the SSD, reducing power consumption and speeding up data retrieval by performing in-SSD database records sorting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sorting operations are performed using host CPU and RAM resources, then data sorting can be accomplished, but power consumption and communication interface utilization increase
Solution Approach 1:
The patent extracts the sorting function from the host CPU and relocates it to a dedicated SSD hardware accelerator. This separation removes the sorting workload from the host system, reducing host power consumption and communication interface utilization while maintaining sorting functionality through specialized hardware in the SSD.
Solution Approach 2:
The patent introduces an embedded reconfigurable SSD hardware accelerator as an intermediary component between the host and the storage medium. This accelerator handles sorting operations locally within the SSD, acting as a mediator that eliminates the need for continuous host CPU intervention and reduces communication overhead through the host interface.
2Productivity
If external sorting processes are used, then data can be sorted, but communication interface utilization increases
Solution Approach 1:
The patent extracts the sorting operation from the host system and implements it within the SSD hardware accelerator. This eliminates the need for data to be transferred back and forth between host and storage, reducing communication interface utilization while maintaining sorting capability through in-SSD processing.
Solution Approach 2:
The SSD hardware accelerator performs sorting operations autonomously without requiring continuous host intervention. The accelerator self-manages the sorting process by directly accessing storage media and generating sorted output, making the system self-sufficient for sorting tasks and reducing host-SSD communication overhead.
3Speed
If sorting is performed without dedicated hardware acceleration, then device complexity is reduced, but data retrieval speed decreases
Solution Approach 1:
The patent introduces an embedded reconfigurable SSD hardware accelerator as an intermediary component that speeds up data retrieval by performing sorting operations locally. This mediator layer between storage media and host provides acceleration without requiring complex changes to the overall system architecture, as it integrates seamlessly within the existing SSD framework.
Solution Approach 2:
The patent replaces the software-based sorting mechanism (which relies on host CPU and RAM) with a dedicated hardware accelerator. This substitution of mechanical/computational approach uses specialized hardware circuits optimized for sorting operations, achieving faster data retrieval speeds while maintaining reasonable device complexity through targeted hardware enhancement rather than system-wide complexity increase.
Data Source
AI summary
A system and associated method for a storage device. In the system and method, a sort command from a host is received. In response to having received the sort command, the storage device is configured for sorting by allocating at least one storage location to store data to be sorted based at least on a) an amount of the data to be sorted and b) an available storage space in a memory of the storage device. In the system and method, an upgradable sorting algorithm is utilized to generate sorted data subsets, and the sorted serialized data subsets are stored based on the at least one storage location indicating where the data to be sorted is to be stored. Page identifiers (IDs) of the sorted data subsets are provided to the host.