Garbage Collection Source Block Selection Using Valid Data Count and Temperature
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional garbage collection schemes in memory sub-systems experience fluctuations in valid data counts (VDC) of garbage collection source blocks, leading to inconsistencies in memory device performance and increased write amplification.
Innovation Solution
A garbage collection management component determines the VDC of each block, identifies blocks satisfying a VDC constraint, and selects a garbage collection source block using a combination of the greedy VDC scheme and the 2D-greedy scheme, based on comparison criteria to minimize VDC deviations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional garbage collection schemes are used, then garbage collection operations can be performed, but valid data count fluctuations occur leading to performance inconsistency and increased write amplification
Solution Approach 1:
The patent changes the selection parameters for garbage collection source blocks by introducing a two-dimensional selection criterion that considers both valid data count and data temperature (access frequency), rather than relying on single-parameter conventional schemes. This parameter change enables more controlled selection that reduces VDC fluctuations and write amplification
Solution Approach 2:
The patent implements feedback mechanisms by monitoring valid data count variations and using this information to adjust subsequent garbage collection source block selections. The system tracks VDC changes and uses this feedback to maintain performance consistency and reduce write amplification in ongoing garbage collection operations
2Device complexity
If single-parameter garbage collection selection is used, then selection process is simple, but valid data count variation amplitude and frequency increase
Solution Approach 1:
The patent transitions from single-parameter selection to two-dimensional selection by introducing data temperature (access frequency) as an additional dimension alongside valid data count. This dimensional expansion creates a more stable selection process that reduces VDC fluctuations while managing complexity through structured multi-criteria evaluation
3Productivity
If greedy VDC scheme is used, then blocks with lowest valid data count are selected, but this causes large deviations in valid data count of source blocks
Solution Approach 1:
The patent modifies the selection parameters by combining valid data count with data temperature metrics, changing from pure VDC-based selection to a hybrid selection model. This parameter change maintains garbage collection efficiency while improving precision in controlling VDC deviations through balanced selection criteria
Solution Approach 2:
The patent introduces dynamic selection that adapts to different block characteristics by considering both VDC and access frequency. The selection process dynamically adjusts based on the combination of these parameters, enabling efficient garbage collection while maintaining precise control over source block VDC variations
Data Source
AI summary
A subset of blocks from a set of blocks of a memory device are identified based on a valid data count constraint. A first block from the subset of blocks is selected based on a valid data count of the first block. A second block from the subset of blocks is selected based on a data temperature of the second block. A comparison of the first block and the second block is performed in accordance with one or more comparison criterion. The first block or the second block is selected as a garbage collection source block based on the comparison. Garbage collection is performed at the garbage collection source block.