Fabric Manager Data Mover Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data mover selection techniques in information handling systems, such as round-robin methods, lead to inefficient data transfers due to memory access time variations depending on the location of data mover devices relative to memory subsystems in NUMA configurations, resulting in slower data transfers.
Innovation Solution
An Information Handling System (IHS) with a fabric manager engine that receives local data mover selection information from multiple computing devices, generates global data mover selection information, and identifies a data mover device with higher priority based on affinity to memory locations for performing data transfers, optimizing data transfer operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If round-robin data mover device selection is used, then data transfer operations are distributed evenly across available data mover devices, but data transfer efficiency deteriorates due to memory access time variations
Solution Approach 1:
The patent applies local quality by selecting data mover devices based on their specific affinity to memory locations rather than treating all data mover devices uniformly. The fabric manager evaluates the relative position of each data mover device to the source and destination memory locations and selects the device with the highest affinity (lowest access time), thereby optimizing local data transfer performance according to spatial characteristics.
2Productivity
If multiple data mover devices are provided in NUMA configurations, then data transfer capacity increases, but selection complexity increases due to varying memory access times
Solution Approach 1:
The patent implements feedback by having the fabric manager continuously monitor and evaluate the relative positions of data mover devices to memory locations. This feedback mechanism allows the system to dynamically select the optimal data mover device based on current spatial relationships, transforming the complex selection problem into a systematic evaluation process that considers memory access time variations.
Solution Approach 2:
The patent applies preliminary action by pre-establishing the fabric manager engine that maintains knowledge of data mover device positions and memory location relationships. This preliminary setup enables the system to quickly determine the optimal data mover device when a data transfer operation is requested, avoiding complex real-time calculations and simplifying the selection process.
Data Source
AI summary
A data mover selection system includes a fabric manager system coupled to computing devices that are coupled to a memory system. The fabric manager system receives respective local data mover selection information from the computing devices that identifies data mover device(s) accessible to those computing device, and generates global data mover selection information that includes each data mover device accessible to the computing devices. When the fabric manager system receives a first data transfer request to transfer data between first and second memory locations in the memory system, it uses the global data mover selection information to identify a data mover device having the highest priority for performing data transfers between the first and second memory locations in the memory system, and transmits a first data transfer instruction to that data mover device to cause that data mover device to perform the first data transfer operation.