Volatile Memory Cold Data Eviction via Refresh Counting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Volatile memory devices, such as DRAMs, face data loss due to leakage currents, requiring frequent refresh operations to maintain data integrity. Existing techniques do not efficiently manage and evict 'cold data' that requires more frequent refreshes, leading to power consumption issues.
Innovation Solution
A memory system that includes a volatile memory device and a non-volatile memory device, where the volatile memory device determines 'cold data' based on a refresh count for each row and evicts this data to the non-volatile memory device, reducing the need for frequent refresh operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If refresh operations are performed frequently to maintain data integrity in volatile memory, then data reliability is improved, but power consumption increases
Solution Approach 1:
The patent segments data into hot data and cold data based on access patterns and refresh counts. By dividing the memory management approach, the system can apply different refresh strategies to different data segments, refreshing cold data less frequently while maintaining overall data integrity.
Solution Approach 2:
The patent performs preliminary classification of data as cold or hot before refresh operations based on refresh counts and access patterns. This preliminary identification allows the system to proactively manage cold data by evicting it to non-volatile memory before it becomes a power consumption burden, rather than continuously refreshing it.
2Reliability
If cold data is continuously refreshed in volatile memory, then data availability is improved, but energy waste increases
Solution Approach 1:
The patent extracts cold data from the volatile memory system by identifying rows with high refresh counts and evicting them to non-volatile memory. This extraction removes the energy-draining refresh operations from the volatile memory subsystem while preserving data availability through the hybrid memory architecture.
Solution Approach 2:
The patent changes the refresh parameter (refresh count) as a criterion for data classification. By monitoring refresh counts and using them to identify cold data, the system dynamically adjusts which data receives frequent refreshes versus which data should be evicted, optimizing energy consumption based on actual refresh behavior.
3Measurement precision
If refresh count monitoring is implemented for each row, then cold data identification accuracy is improved, but device complexity increases
Solution Approach 1:
The patent implements a self-service mechanism where the refresh operation circuit automatically increments refresh counts for each row without requiring external intervention. The table management circuit then autonomously monitors these counts and triggers eviction decisions based on predetermined thresholds, reducing the need for complex external control logic.
Solution Approach 2:
The patent establishes a feedback loop where refresh operation results (refresh counts) are fed back to the table management circuit, which uses this information to make intelligent eviction decisions. This feedback mechanism enables accurate cold data identification while keeping the overall system manageable through structured information flow.
Data Source
AI summary
A memory system may include: a non-volatile memory device, a volatile memory device suitable for defining, as cold data, data stored in a word line, on which a refresh operation is performed a number of times greater than a reference number among a plurality of word lines coupled to a volatile memory cell array, and evicting the cold data, and a controller suitable for controlling operations of the volatile memory device and the non-volatile memory device, and storing the evicted cold data into the non-volatile memory device.