Memory Controller Access Tracking for Tiered Data Placement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing tier-aware software implementations that monitor access patterns across memory tiers primarily at the host system reduce software performance by consuming resources that could be used for other services, and lower sampling rates to mitigate this impact decrease the accuracy of access pattern tracking.
Innovation Solution
Implementing access pattern monitoring primarily at the memory system, using a memory system controller to track access patterns by calculating hash values, incrementing counters, and storing frequently accessed page indices in a FIFO buffer, which provides accurate access patterns with minimal impact on host system performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If access pattern monitoring is implemented primarily at the host system, then access pattern tracking can be performed, but software performance is reduced due to resource consumption
Solution Approach 1:
The patent introduces a memory system controller as an intermediary component that performs access pattern monitoring. This mediator handles the monitoring tasks between the host system and memory devices, preventing the host system from directly consuming resources for monitoring while maintaining accurate tracking capability.
Solution Approach 2:
The memory system is made self-sufficient by enabling it to perform access pattern monitoring independently through its own controller. The memory system monitors its own access patterns without requiring host system resources, allowing the host to focus on its primary computational tasks while the memory system autonomously tracks access patterns for optimization purposes.
2Productivity
If a lower sampling rate is used to mitigate performance impact, then software performance is improved, but access pattern tracking accuracy is reduced
Solution Approach 1:
The memory system controller acts as an intermediary that can monitor access patterns at full resolution without impacting host system performance. By offloading the monitoring function to the intermediary controller, the system achieves both high sampling accuracy and maintained host performance, eliminating the need to compromise between these two factors.
3Measurement precision
If access pattern monitoring is performed at the host system, then access patterns can be tracked, but host system resources are consumed for other services
Solution Approach 1:
The access pattern monitoring function is extracted from the host system and relocated to the memory system controller. This separation removes the monitoring burden from the host system, allowing host resources to be dedicated to other services while the extracted monitoring function operates independently in the memory subsystem.
Solution Approach 2:
The memory system is empowered to monitor its own access patterns autonomously through its controller, without requiring host system intervention or resource allocation. This self-service capability enables accurate tracking while keeping the host system simple and focused on its primary functions.
Data Source
AI summary
Methods, systems, and devices for access pattern tracking are described. An access request associated with accessing first data stored in memory may be received. Based on receiving the access request, a hash value based on a page index indicated in the access request may be calculated. Based on the hash value, a counter associated with the hash value may be incremented. Based on incrementing the counter, an access pattern associated with access data in the memory may be indicated.