Data Mover Engine for Memory Fabric Tiering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional memory tiering systems in information handling systems require network protocols and dedicated software stacks to move data between memory subsystems, which increases processing overhead and complexity.
Innovation Solution
A data mover engine within the information handling system that uses a memory fabric to dynamically move data between memory subsystems based on access frequency, optimizing memory access by mapping compute contexts to memory subsystems with appropriate performance characteristics without relying on network protocols or dedicated software stacks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional memory tiering operations use network protocols and dedicated software stacks, then data can be moved between memory subsystems, but processing overhead and system complexity increase
Solution Approach 1:
The patent extracts the memory tiering functionality from the compute host and places it in a dedicated data mover device. This separation removes the complexity of network protocols and software stacks from the host system while maintaining data movement capability through specialized hardware designed for this specific function.
Solution Approach 2:
The data mover device acts as an intermediary between the compute host and memory subsystems. It handles all data movement operations between different memory tiers without requiring the host to directly manage complex protocols, thus simplifying the host system while enabling sophisticated memory tiering.
2Productivity
If conventional memory tiering uses dedicated software stacks executed by the processing system, then memory tiering operations can be performed, but processing system cycles are consumed that could be used for higher value processing operations
Solution Approach 1:
The patent extracts memory tiering operations from the processing system and relocates them to a dedicated data mover device. This extraction frees processing system cycles for higher-value computations while the data mover handles memory management tasks independently through hardware-based operations.
Solution Approach 2:
The data mover device performs memory tiering operations autonomously without requiring processing system intervention. It independently manages data movement between memory subsystems, allowing the processing system to focus on computational tasks while the data mover self-manages memory optimization.
3Quantity of substance
If data is stored in memory subsystems outside the server device, then more room is available for regularly used data, but memory access latency increases for frequently accessed data
Solution Approach 1:
The patent implements dynamic memory tiering that automatically adjusts data placement based on access patterns. Frequently accessed data is dynamically moved to lower-latency memory subsystems, while less frequently accessed data resides in external memory, thus optimizing both capacity utilization and access latency based on real-time workload conditions.
Solution Approach 2:
The system uses feedback from data access patterns to make intelligent decisions about data placement. By monitoring access frequency and performance metrics, the data mover dynamically adjusts which data resides in which memory tier, ensuring that hot data is positioned for fast access while cold data utilizes external capacity.
Data Source
AI summary
A memory tiering system includes a data mover device coupling a memory fabric to a processing system. At each of a plurality of different times, the data mover device receives a data access request for data from a computer context provided by the processing system, retrieves the data based on a compute-context-memory-fabric mapping that maps the compute context to the first memory subsystem, and provides the data to the processing system for use with the computer context. If the data mover device determines that the data has been retrieved and provisioning for use with the compute context above a memory tiering frequency, it moves the data from the first memory subsystem to a second memory subsystem in the memory fabric that includes higher performance memory characteristics, and causes the compute-context-memory-fabric mapping to be modified to provide a modified compute-context-memory-fabric mapping that maps the compute context to the second memory subsystem.