Memory Controller Counting Bloom Filter for Row Hammer Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory devices are susceptible to row hammer events and endurance degrading events due to shrinking memory sizes, with current detection methods being either inaccurate or requiring excessive resources, and existing row hammer detection algorithms can be exploited or are impractical.
Innovation Solution
Integration of a counting bloom filter into the memory controller to detect row hammer events using less memory and power, with implementations at global, channel, and memory bank levels, ensuring accurate detection without false negatives and reducing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing row hammer detection algorithms are used, then detection capability is provided, but memory resources and power consumption increase excessively
Solution Approach 1:
The patent extracts only the essential detection functionality needed for row hammer detection and implements it through a bloom filter data structure. Instead of using comprehensive detection algorithms that require extensive memory, the bloom filter captures the critical detection capability while using minimal memory resources by storing only hashed representations of accessed row addresses.
Solution Approach 2:
The patent inverts the traditional approach by using a probabilistic data structure (bloom filter) that is designed to minimize false negatives rather than minimizing false positives. This inversion allows the system to detect row hammer events with high reliability while using far less memory than traditional detection methods that store complete address histories.
2Reliability
If existing row hammer detection algorithms are used, then detection capability is provided, but power consumption increases excessively
Solution Approach 1:
The patent extracts only the essential detection functionality needed for row hammer detection and implements it through a bloom filter data structure. Instead of using comprehensive detection algorithms that require extensive memory, the bloom filter captures the critical detection capability while using minimal memory resources by storing only hashed representations of accessed row addresses.
Solution Approach 2:
The patent uses hashing to create compact representations (copies) of the accessed row addresses. Instead of storing the actual addresses which would consume significant memory and power, the bloom filter stores hashed values that are much smaller in size, thereby reducing both memory usage and the power required to maintain and search the detection structure.
3Quantity of substance
If memory size is reduced to meet scaling requirements, then device density improves, but susceptibility to row hammer events increases
Solution Approach 1:
The patent introduces a bloom filter as an intermediary layer between the memory access operations and the physical memory cells. This intermediary structure monitors and tracks accessed row addresses through hashing, enabling the detection of row hammer patterns without requiring changes to the underlying memory cell structure or increasing physical memory size, thus maintaining device density while providing protection.
Data Source
AI summary
Systems, apparatuses, and methods related to bloom filter implementation into a controller are described. A memory device is coupled to a memory controller. The memory controller is configured to implement a counting bloom filter, increment the counting bloom filter in response to a row activate command of the memory device, determine whether a value of the counting bloom filter exceeds a threshold value, and perform an action in response to the value exceeding the threshold value.