Network Storage Gateway with Cache SSDs for CPU and Memory Bottleneck Relief
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Converged storage systems face CPU bottlenecks due to the central processing unit handling both computing and storage operations, and memory bottlenecks due to increased data processing demands, which limit scalability and introduce latency issues.
Innovation Solution
Implementing a disaggregated infrastructure with cache SSDs equipped with storage processing units (SPUs) and general data processing units (DPUs) to offload storage management and application processing from the CPU, utilizing Ethernet and PCIe communication, and incorporating Key-Value SSDs for efficient data handling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If storage devices are increased in size and data storage capacity is expanded, then storage capability is improved, but data processing efficiency deteriorates due to CPU and memory bottlenecks
Solution Approach 1:
The patent segments the monolithic storage system into distributed storage nodes, each with local cache memory and processing capabilities. This segmentation allows parallel data processing across multiple nodes, eliminating the CPU and memory bottlenecks that would otherwise occur when a single central processor handles all storage operations for large-capacity storage devices.
Solution Approach 2:
The patent introduces cache memory as an intermediary between the high-capacity storage devices and the CPU. This intermediary layer buffers data transfers, allowing the CPU to continue processing other tasks while data is being read or written to/from large storage devices, thereby maintaining data processing efficiency despite increased storage capacity.
2Adaptability or versatility
If computing server and storage server are converged into a single unit, then system integration is improved, but CPU bottleneck is worsened due to centralized processing demands
Solution Approach 1:
The patent segments the converged system into distributed storage nodes that each have local processing capabilities and cache memory. This segmentation distributes the processing load across multiple independent units rather than concentrating it in a single CPU, thereby maintaining system integration while avoiding CPU bottlenecks.
Solution Approach 2:
Each storage node in the patent is equipped with its own cache memory and processing logic, allowing it to independently handle data requests and preprocessing tasks. This self-service capability reduces the burden on centralized CPU resources while maintaining the integrated nature of the converged system.
3Quantity of substance
If cache memory size is increased to handle larger data sets, then memory capacity is improved, but memory bottleneck is worsened due to bandwidth limitations
Solution Approach 1:
The patent segments the memory system into distributed cache memory across multiple storage nodes. This segmentation allows parallel data access across multiple memory channels, effectively increasing both memory capacity and aggregate bandwidth. The distributed architecture prevents memory bottlenecks by allowing simultaneous data operations across multiple nodes rather than through a single memory interface.
Data Source
AI summary
A Solid State Drive (SSD) is disclosed. The SSD may include ports to receive requests from a host and to send requests to a second storage device. The SSD may include flash storage for data. An SSD controller may process the requests received from the host and generate the requests sent to the second storage device. The SSD may act as a cache for the second storage device.