Row Hammer Counting with Ping-Pong Tables for Zero False Negatives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional row hammer detection techniques are either impractical due to high resource requirements or imperfect in preventing data corruption, and row hammer attacks pose a significant risk in memory devices, especially in hyperscale datacenters where they can silently corrupt other users' data.
Innovation Solution
Implementing a row hammer detector with row access counters at the per-bank or per-channel level in a memory controller, using aliasing to reduce the number of counters needed and enabling perfect tracking of row hammer attacks in an energy and space-efficient manner.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional row hammer detection techniques are used, then data corruption can be detected, but the resource requirements become excessively high making them impractical
Solution Approach 1:
The patent segments the row hammer detection system by implementing counters at the per-bank or per-channel level rather than tracking every individual row globally. This segmentation reduces the total number of counters needed while maintaining detection effectiveness, as row hammer attacks typically concentrate on specific banks or channels.
Solution Approach 2:
The patent merges multiple row tracking functions into a unified counter system organized by bank or channel. Instead of maintaining separate counters for each row, the system combines them into shared counters that track aggregate row activation patterns, significantly reducing resource requirements while preserving detection capability.
2Reliability
If conventional row hammer detection techniques are used, then detection can be performed, but false negatives occur allowing data corruption to slip through
Solution Approach 1:
The patent implements feedback mechanisms where the counter system continuously monitors row activation patterns and provides real-time detection signals. When a counter exceeds the row hammer threshold, the system generates detection feedback that triggers mitigation responses, ensuring complete detection without false negatives through continuous monitoring and immediate response.
3Reliability
If perfect row tracking is implemented, then all row hammer attacks can be detected, but the implementation requires excessive memory and power resources
Solution Approach 1:
The patent applies partial action by implementing row hammer detection at the per-bank or per-channel level rather than attempting to track every individual row globally. This partial implementation achieves sufficient detection coverage for practical purposes while consuming significantly less power and memory resources than a complete global tracking system would require.
4Reliability
If perfect row tracking is implemented, then all row hammer attacks can be detected, but the implementation requires excessive memory resources
Solution Approach 1:
The patent segments the detection system into per-bank or per-channel counters, which dramatically reduces the total memory footprint. Instead of allocating memory for every individual row across the entire memory array, the system uses a small number of shared counters organized by bank or channel, achieving perfect tracking within each segment with minimal memory resources.
Data Source
AI summary
Systems and methods for finite time counting period counting of infinite data streams is presented. In particular example systems and methods enable counting row accesses to a memory media device over predetermined time intervals in order to deterministically detect row hammer attacks on the memory media device. Example embodiments use two identical tables that are reset at times offset in relation to each other in a ping-pong manner in order to ensure that there exists no false negative detections. The counting techniques described in this disclosure can be used in various types of row hammer mitigation techniques and can be implemented in content addressable memory or another type of memory. The mitigation may be implemented on a per-bank basis, per-channel basis or per-memory media device basis. The memory media device may be a dynamic random access memory type device.