Bridging Device Memory Mapping for Low-Latency CXL Resource Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing resource management methods in data centers using Compute Express Link (CXL) technology result in increased internal memory access latency due to centralized management by a master control host, degrading system performance.
Innovation Solution
A resource management method and apparatus utilizing a bridging device with an instruction parsing module, internal memory mapping table, and switch matrix module to directly process host commands for memory allocation and release, enabling and disabling channels based on the internal memory mapping table and switch matrix module.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a master control host uniformly manages and allocates all internal memory resources, then resource management is centralized and simplified, but internal memory access latency increases and system performance degrades
Solution Approach 1:
The patent segments the centralized resource management function into distributed management nodes. Each host or storage device is equipped with independent resource management capabilities, allowing local decision-making for resource allocation and release. This segmentation eliminates the single-point bottleneck of the master control host, reducing access latency while maintaining manageable complexity through modular design.
Solution Approach 2:
The patent introduces an intermediary resource management module that sits between hosts and internal memory devices. This intermediary handles resource allocation requests locally, acting as a mediator that eliminates the need for all hosts to communicate directly with a central controller. The intermediary maintains resource mapping tables and manages channel switching, thereby reducing access latency while keeping the system manageable.
2Reliability
If all hosts send requests to the master control host to acquire or release resources, then centralized control is maintained, but resource management processing time increases
Solution Approach 1:
The patent divides the centralized resource management function into multiple distributed management nodes. Each node can independently handle resource allocation and release requests for its local resources, eliminating the need for all hosts to communicate with a single master control host. This segmentation maintains control reliability through distributed consensus while dramatically improving resource allocation speed by eliminating communication bottlenecks.
Solution Approach 2:
The patent implements preliminary action by pre-configuring resource mapping tables and establishing resource management capabilities at each host or storage device before runtime. Resource allocation decisions are made based on pre-established policies and local resource states, eliminating the need for real-time communication with a central controller for every allocation request. This improves productivity while maintaining reliability through pre-planned resource management strategies.
Data Source
AI summary
Provided are a resource management method and apparatus, an electronic device, and a non-volatile readable storage medium. The method includes: determining, in response to a target command sent by a host, a command type corresponding to the target command by a bridging device, in a case that the target command is an internal memory application command, a first target resource allocated to the host based on an internal memory mapping table, and enabling a first channel between the first target resource and the host; and determining, in a case that the target command is an internal memory release command, a second target resource indicated by the internal memory release command, and disabling a second channel between the second target resource and the host.